"
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"# Visualization of clean/noisy/spike/drift\n",
"\n",
"def plot_demo_example(demo, feature_names):\n",
" clean_orig = demo[\"clean_orig\"]\n",
" noisy_orig = demo[\"noisy_orig\"]\n",
" spike_orig = demo[\"spike_orig\"]\n",
" drift_orig = demo[\"drift_orig\"]\n",
"\n",
" clean_std = demo[\"clean_std\"]\n",
" spike_std = demo[\"spike_std\"]\n",
" drift_std = demo[\"drift_std\"]\n",
"\n",
" spike_feat = demo[\"spike_feat\"]\n",
" drift_feat = demo[\"drift_feat\"]\n",
" spike_t0 = demo[\"spike_t0\"]\n",
" drift_t0 = demo[\"drift_t0\"]\n",
"\n",
" t_hours = np.arange(clean_orig.shape[0]) * (10.0 / 60.0)\n",
"\n",
" fig, axes = plt.subplots(2, 2, figsize=(16, 9), dpi=140)\n",
"\n",
" # 1) clean vs noisy\n",
" ax = axes[0, 0]\n",
" ax.plot(t_hours, clean_orig[:, spike_feat], linewidth=2, label=\"clean\")\n",
" ax.plot(t_hours, noisy_orig[:, spike_feat], linewidth=1.8, alpha=0.9, label=\"noisy input\")\n",
" ax.set_title(f\"Clean vs noisy\\nFeature: {feature_names[spike_feat]}\")\n",
" ax.set_xlabel(\"Hours in window\")\n",
" ax.set_ylabel(\"Value\")\n",
" ax.grid(True, alpha=0.3)\n",
" ax.legend()\n",
"\n",
" # 2) clean vs spike\n",
" ax = axes[0, 1]\n",
" ax.plot(t_hours, clean_orig[:, spike_feat], linewidth=2, alpha=0.9, color=\"blue\",label=\"clean\")\n",
" ax.plot(t_hours, spike_orig[:, spike_feat], linewidth=2, alpha=0.9, color=\"orange\", label=\"spike injected\")\n",
" ax.axvline(t_hours[spike_t0], color=\"red\", linestyle=\"--\", linewidth=1.5, label=\"spike time\")\n",
" ax.scatter(t_hours[spike_t0], clean_orig[spike_t0, spike_feat], color=\"black\", s=45, zorder=5)\n",
" ax.scatter(t_hours[spike_t0], spike_orig[spike_t0, spike_feat], color=\"red\", s=55, zorder=6)\n",
" ax.set_title(f\"Spike anomaly\\nFeature: {feature_names[spike_feat]}\")\n",
" ax.set_xlabel(\"Hours in window\")\n",
" ax.set_ylabel(\"Value\")\n",
" ax.grid(True, alpha=0.3)\n",
" ax.legend()\n",
"\n",
" # 3) clean vs drift\n",
" ax = axes[1, 0]\n",
" ax.plot(t_hours, clean_orig[:, drift_feat], linewidth=2, alpha=0.9, label=\"clean\")\n",
" ax.plot(t_hours, drift_orig[:, drift_feat], linewidth=2, alpha=0.9, label=\"drift injected\")\n",
" ax.axvline(t_hours[drift_t0], color=\"red\", linestyle=\"--\", linewidth=1.5, label=\"drift start\")\n",
" ax.set_title(f\"Drift anomaly\\nFeature: {feature_names[drift_feat]}\")\n",
" ax.set_xlabel(\"Hours in window\")\n",
" ax.set_ylabel(\"Value\")\n",
" ax.grid(True, alpha=0.3)\n",
" ax.legend()\n",
"\n",
" # 4) standardized-unit heatmaps (better than original units for comparison)\n",
" ax = axes[1, 1]\n",
" spike_diff_std = np.abs(spike_std - clean_std).T\n",
" drift_diff_std = np.abs(drift_std - clean_std).T\n",
"\n",
" # stack them with a blank separator row for readability\n",
" sep = np.full((1, spike_diff_std.shape[1]), np.nan)\n",
" combined = np.vstack([spike_diff_std, sep, drift_diff_std])\n",
"\n",
" im = ax.imshow(combined, aspect=\"auto\", origin=\"lower\")\n",
" ax.set_title(\"Absolute change in standardized units\\nTop = spike, bottom = drift\")\n",
" ax.set_xlabel(\"Timestep\")\n",
" ax.set_ylabel(\"Feature\")\n",
" n_feat = len(feature_names)\n",
"\n",
" yticks = list(range(n_feat)) + list(range(n_feat + 1, 2 * n_feat + 1))\n",
" ylabels = feature_names + feature_names\n",
" ax.set_yticks(yticks)\n",
" ax.set_yticklabels(ylabels, fontsize=8)\n",
"\n",
" plt.colorbar(im, ax=ax, fraction=0.046, pad=0.04)\n",
"\n",
" fig.suptitle(\"Illustration of clean, noisy, spike, and drift windows\", fontsize=16, y=0.98)\n",
" plt.tight_layout()\n",
" plt.show()\n",
"\n",
"plot_demo_example(demo, feature_names)\n"
]
},
{
"cell_type": "markdown",
"id": "d4d14d2a",
"metadata": {},
"source": [
"## 6. Model definitions\n",
"\n",
"Two autoencoders are used:\n",
"\n",
"- **1D‑CNN AE:** learns local temporal motifs with parallel computation\n",
"- **LSTM AE:** learns temporal dependencies using recurrent state\n",
"\n",
"Both models map `(L, F) → (L, F)` and are trained with MAE reconstruction loss.\n"
]
},
{
"cell_type": "code",
"execution_count": 13,
"id": "b5aa823e",
"metadata": {},
"outputs": [],
"source": [
"def build_cnn_autoencoder(window: int, n_features: int, latent: int = 32) -> keras.Model:\n",
" # CNN autoencoder for window reconstruction.\n",
" # Note: window must be divisible by 4 due to two downsampling steps (pool_size=2 twice).\n",
" if window % 4 != 0:\n",
" raise ValueError(\"window must be divisible by 4 for this CNN AE architecture.\")\n",
"\n",
" inp = layers.Input(shape=(window, n_features))\n",
"\n",
" x = layers.Conv1D(32, 3, padding=\"same\", activation=\"relu\")(inp)\n",
" x = layers.MaxPooling1D(pool_size=2)(x)\n",
" x = layers.Conv1D(64, 3, padding=\"same\", activation=\"relu\")(x)\n",
" x = layers.MaxPooling1D(pool_size=2)(x)\n",
"\n",
" # Dense bottleneck: fixed-size latent vector for a controlled CNN vs LSTM comparison.\n",
" x = layers.Flatten()(x)\n",
" z = layers.Dense(latent, activation=\"relu\")(x)\n",
"\n",
" x = layers.Dense((window // 4) * 64, activation=\"relu\")(z)\n",
" x = layers.Reshape((window // 4, 64))(x)\n",
"\n",
" x = layers.UpSampling1D(size=2)(x)\n",
" x = layers.Conv1D(64, 3, padding=\"same\", activation=\"relu\")(x)\n",
" x = layers.UpSampling1D(size=2)(x)\n",
"\n",
" out = layers.Conv1D(n_features, 3, padding=\"same\", activation=None)(x)\n",
"\n",
" model = keras.Model(inp, out, name=\"cnn_autoencoder\")\n",
" model.compile(optimizer=keras.optimizers.Adam(1e-3), loss=\"mae\")\n",
" return model\n",
"\n",
"def build_lstm_autoencoder(window: int, n_features: int, latent: int = 32) -> keras.Model:\n",
" # LSTM autoencoder for window reconstruction.\n",
" inp = layers.Input(shape=(window, n_features))\n",
"\n",
" x = layers.LSTM(latent, return_sequences=False)(inp)\n",
" x = layers.RepeatVector(window)(x)\n",
" x = layers.LSTM(latent, return_sequences=True)(x)\n",
"\n",
" out = layers.TimeDistributed(layers.Dense(n_features))(x)\n",
"\n",
" model = keras.Model(inp, out, name=\"lstm_autoencoder\")\n",
" model.compile(optimizer=keras.optimizers.Adam(1e-3), loss=\"mae\")\n",
" return model\n"
]
},
{
"cell_type": "markdown",
"id": "bf7acccf",
"metadata": {},
"source": [
"## 7. Training utilities\n",
"\n",
"Training uses:\n",
"- early stopping on validation loss\n",
"- learning‑rate reduction when validation loss plateaus\n",
"\n",
"These controls limit wasted compute and stabilize optimization.\n"
]
},
{
"cell_type": "code",
"execution_count": 14,
"id": "220738cd",
"metadata": {},
"outputs": [],
"source": [
"def train_autoencoder(model: keras.Model, ds_train: tf.data.Dataset, ds_val: tf.data.Dataset, epochs: int = 30):\n",
" callbacks = [\n",
" keras.callbacks.EarlyStopping(monitor=\"val_loss\", patience=5, restore_best_weights=True),\n",
" keras.callbacks.ReduceLROnPlateau(monitor=\"val_loss\", factor=0.5, patience=2, min_lr=1e-5),\n",
" ]\n",
" t0 = time.time()\n",
" history = model.fit(ds_train, validation_data=ds_val, epochs=epochs, callbacks=callbacks, verbose=2)\n",
" dt = time.time() - t0\n",
" return history, dt"
]
},
{
"cell_type": "markdown",
"id": "ddef8b7d",
"metadata": {},
"source": [
"## 7.0 Evaluation utilities\n",
"\n",
"Helper functions used by both the lightweight justification runs and the main evaluation.\n"
]
},
{
"cell_type": "code",
"execution_count": 15,
"id": "af4c154a",
"metadata": {},
"outputs": [],
"source": [
"def window_scores_mean(model: keras.Model, ds_windows: tf.data.Dataset) -> np.ndarray:\n",
" # Mean absolute reconstruction error per window: (B,L,F) -> (B,)\n",
" scores = []\n",
" for batch in ds_windows:\n",
" pred = model(batch, training=False)\n",
" err = tf.abs(batch - pred) # (B,L,F)\n",
" s = tf.reduce_mean(err, axis=[1, 2]) # (B,)\n",
" scores.append(s.numpy())\n",
" return np.concatenate(scores, axis=0)\n",
"\n",
"def window_scores_max(model: keras.Model, ds_windows: tf.data.Dataset) -> np.ndarray:\n",
" # Max absolute reconstruction error per window (spike-sensitive): (B,L,F) -> (B,)\n",
" scores = []\n",
" for batch in ds_windows:\n",
" pred = model(batch, training=False)\n",
" err = tf.abs(batch - pred) # (B,L,F)\n",
" s = tf.reduce_max(err, axis=[1, 2]) # (B,)\n",
" scores.append(s.numpy())\n",
" return np.concatenate(scores, axis=0)\n",
"\n",
"def window_scores_topk(model: keras.Model, ds_windows: tf.data.Dataset, k: int = 20) -> np.ndarray:\n",
" # Mean of top-k absolute errors per window. Often balances drift and spike detection.\n",
" scores = []\n",
" for batch in ds_windows:\n",
" pred = model(batch, training=False)\n",
" err = tf.abs(batch - pred) # (B,L,F)\n",
" flat = tf.reshape(err, (tf.shape(err)[0], -1)) # (B, L*F)\n",
" topk = tf.math.top_k(flat, k=k).values\n",
" s = tf.reduce_mean(topk, axis=1) # (B,)\n",
" scores.append(s.numpy())\n",
" return np.concatenate(scores, axis=0)\n",
"\n",
"def percentile_threshold(scores: np.ndarray, q: float = 99.0) -> float:\n",
" return float(np.percentile(scores, q))"
]
},
{
"cell_type": "code",
"execution_count": 16,
"id": "6cfc45ff",
"metadata": {},
"outputs": [],
"source": [
"def _score_from_error(err: tf.Tensor, mode: str = \"mean\", topk: int = 20) -> np.ndarray:\n",
" # err has shape (B, L, F). Returns (B,) score.\n",
" if mode == \"mean\":\n",
" s = tf.reduce_mean(err, axis=[1, 2])\n",
" elif mode == \"max\":\n",
" s = tf.reduce_max(err, axis=[1, 2])\n",
" elif mode == \"topk\":\n",
" flat = tf.reshape(err, (tf.shape(err)[0], -1))\n",
" s = tf.reduce_mean(tf.math.top_k(flat, k=topk).values, axis=1)\n",
" else:\n",
" raise ValueError(\"mode must be one of: 'mean', 'max', 'topk'\")\n",
" return s.numpy()\n",
"\n",
"def eval_on_injected(model: keras.Model, ds_injected: tf.data.Dataset, threshold: float, score_mode: str = \"mean\", topk: int = 20):\n",
" # Evaluates window-level anomaly detection on injected data.\n",
" y_true = []\n",
" y_score = []\n",
" y_pred = []\n",
"\n",
" for x_batch, y_batch, _t0 in ds_injected:\n",
" pred = model(x_batch, training=False)\n",
" err = tf.abs(x_batch - pred) # (B,L,F)\n",
" score = _score_from_error(err, mode=score_mode, topk=topk) # (B,)\n",
" y = y_batch.numpy().astype(int)\n",
"\n",
" y_true.append(y)\n",
" y_score.append(score)\n",
" y_pred.append((score > threshold).astype(int))\n",
"\n",
" y_true = np.concatenate(y_true)\n",
" y_score = np.concatenate(y_score)\n",
" y_pred = np.concatenate(y_pred)\n",
"\n",
" ap = average_precision_score(y_true, y_score) if len(np.unique(y_true)) > 1 else np.nan\n",
" p, r, f1, _ = precision_recall_fscore_support(y_true, y_pred, average=\"binary\", zero_division=0)\n",
" cm = confusion_matrix(y_true, y_pred, labels=[0, 1]) # [[TN,FP],[FN,TP]]\n",
"\n",
" return {\"ap\": float(ap), \"precision\": float(p), \"recall\": float(r), \"f1\": float(f1), \"cm\": cm}\n",
"\n",
"def plot_confusion_matrix(cm: np.ndarray, title: str, cmap: str = \"Blues\"):\n",
" \"\"\"Plot a 2×2 confusion matrix with better contrast.\"\"\"\n",
" labels = [\"Normal (0)\", \"Anomaly (1)\"]\n",
"\n",
" fig, ax = plt.subplots(figsize=(5.0, 4.0), dpi=140)\n",
" im = ax.imshow(cm, interpolation=\"nearest\", cmap=cmap)\n",
" ax.set_title(title)\n",
" ax.set_xlabel(\"Predicted\")\n",
" ax.set_ylabel(\"True\")\n",
"\n",
" ax.set_xticks([0, 1])\n",
" ax.set_yticks([0, 1])\n",
" ax.set_xticklabels(labels, rotation=25, ha=\"right\")\n",
" ax.set_yticklabels(labels)\n",
"\n",
" # Better text contrast based on normalized cell intensity\n",
" cm_float = cm.astype(float)\n",
" vmin, vmax = cm_float.min(), cm_float.max()\n",
" denom = (vmax - vmin) if vmax > vmin else 1.0\n",
"\n",
" for (i, j), v in np.ndenumerate(cm):\n",
" norm_v = (float(v) - vmin) / denom\n",
" color = \"white\" if norm_v > 0.55 else \"black\"\n",
" ax.text(j, i, f\"{v}\", ha=\"center\", va=\"center\", fontsize=10, fontweight=\"bold\", color=color)\n",
"\n",
" cbar = fig.colorbar(im, ax=ax, fraction=0.046, pad=0.04)\n",
" cbar.ax.tick_params(labelsize=8)\n",
"\n",
" fig.tight_layout()\n",
" plt.show()\n",
"\n",
"\n",
"\n",
"# Evaluation: drift is spread across many timesteps → mean score works well.\n",
"# Spike is localized → max (or top-k) score is more sensitive than mean."
]
},
{
"cell_type": "code",
"execution_count": 17,
"id": "4eadb24c",
"metadata": {},
"outputs": [],
"source": [
"# Within-window drift detection delay utilities\n",
"\n",
"def timestep_scores(model: keras.Model, x_batch: tf.Tensor) -> np.ndarray:\n",
" \"\"\"Per-timestep reconstruction score (mean absolute error across features).\n",
" Input: (B, L, F)\n",
" Output: (B, L)\n",
" \"\"\"\n",
" pred = model(x_batch, training=False)\n",
" err = tf.abs(x_batch - pred) # (B,L,F)\n",
" s_t = tf.reduce_mean(err, axis=2) # (B,L)\n",
" return s_t.numpy()\n",
"\n",
"def timestep_threshold_from_val(model: keras.Model, ds_val_windows: tf.data.Dataset, q: float = 99.0) -> float:\n",
" \"\"\"Percentile threshold for per-timestep scores computed on clean validation windows.\"\"\"\n",
" all_scores = []\n",
" for x in ds_val_windows:\n",
" s_t = timestep_scores(model, x) # (B,L)\n",
" all_scores.append(s_t.reshape(-1))\n",
" all_scores = np.concatenate(all_scores)\n",
" return float(np.percentile(all_scores, q))\n",
"\n",
"def detection_delay_stats(model: keras.Model, ds_injected: tf.data.Dataset, timestep_threshold: float):\n",
" \"\"\"Delay stats for drift-injected windows (x, y, t0).\n",
" Delay = first timestep where score > threshold minus drift start t0.\n",
" Returns stats in timesteps.\n",
" \"\"\"\n",
" delays = []\n",
" for x_batch, y_batch, t0_batch in ds_injected:\n",
" s_t = timestep_scores(model, x_batch) # (B,L)\n",
" t0 = t0_batch.numpy()\n",
" y = y_batch.numpy()\n",
"\n",
" for i in range(len(y)):\n",
" if y[i] == 0:\n",
" continue\n",
" start = int(t0[i])\n",
" hits = np.where(s_t[i] > timestep_threshold)[0]\n",
" if hits.size == 0:\n",
" continue\n",
" delays.append(max(0, int(hits[0]) - start))\n",
"\n",
" delays = np.array(delays, dtype=float)\n",
" if delays.size == 0:\n",
" return {\"n\": 0, \"mean\": np.nan, \"median\": np.nan, \"p90\": np.nan}\n",
" return {\n",
" \"n\": int(delays.size),\n",
" \"mean\": float(np.mean(delays)),\n",
" \"median\": float(np.median(delays)),\n",
" \"p90\": float(np.percentile(delays, 90)),\n",
" }"
]
},
{
"cell_type": "markdown",
"id": "31d61246",
"metadata": {},
"source": [
"## 7.1 Lightweight justification runs\n",
"\n",
"The final experiment is trained once with `WINDOW=144` and `SHIFT_TRAIN=6` (denoising AE).\n",
"This section provides small, CPU‑safe checks to justify those choices.\n"
]
},
{
"cell_type": "code",
"execution_count": 18,
"id": "c3da8aeb",
"metadata": {},
"outputs": [],
"source": [
"\n",
"def run_one_setting(\n",
" window: int,\n",
" model_type: str,\n",
" latent: int = 32,\n",
" epochs: int = 10,\n",
" shift_train: int = 1,\n",
" shift_eval: int = 1,\n",
" limit_val_batches: int | None = None,\n",
" limit_test_batches: int | None = None,\n",
" compute_delay: bool = False,\n",
"):\n",
" \"\"\"Runs one configuration and returns one row (dict) for summary tables.\n",
"\n",
" This helper is used for lightweight justification runs. For the final run,\n",
" the notebook trains the chosen models once using WINDOW=144 and SHIFT_TRAIN=6.\n",
"\n",
" Parameters\n",
" ----------\n",
" limit_val_batches, limit_test_batches:\n",
" If set, only a subset of validation/test batches are used (faster, approximate).\n",
" compute_delay:\n",
" If True, also compute within-window drift detection delay (slower).\n",
" \"\"\"\n",
"\n",
" # Window datasets (train stride controls redundancy; eval stride fixed)\n",
" ds_tr = make_window_dataset(X_train, window, shift_train, BATCH_SIZE, shuffle=True)\n",
" ds_vl = make_window_dataset(X_val, window, shift_eval, BATCH_SIZE, shuffle=False)\n",
" ds_te = make_window_dataset(X_test, window, shift_eval, BATCH_SIZE, shuffle=False)\n",
"\n",
" if limit_val_batches is not None:\n",
" ds_vl = ds_vl.take(limit_val_batches)\n",
" if limit_test_batches is not None:\n",
" ds_te = ds_te.take(limit_test_batches)\n",
"\n",
" # DAE training view: (x + noise) -> x\n",
" ds_tr_fit = ds_tr.enumerate().map(lambda i, x: add_gaussian_noise(i, x), num_parallel_calls=tf.data.AUTOTUNE).prefetch(tf.data.AUTOTUNE)\n",
" ds_vl_fit = ds_vl.map(lambda x: (x, x), num_parallel_calls=tf.data.AUTOTUNE).prefetch(tf.data.AUTOTUNE)\n",
"\n",
" # Build model\n",
" if model_type == \"cnn\":\n",
" model = build_cnn_autoencoder(window, F, latent=latent)\n",
" elif model_type == \"lstm\":\n",
" model = build_lstm_autoencoder(window, F, latent=latent)\n",
" else:\n",
" raise ValueError(\"model_type must be 'cnn' or 'lstm'\")\n",
"\n",
" # Train\n",
" hist, t_train = train_autoencoder(model, ds_tr_fit, ds_vl_fit, epochs=epochs)\n",
"\n",
" # Thresholds from clean validation windows\n",
" thr_mean = percentile_threshold(window_scores_mean(model, ds_vl), 99.0) # drift\n",
" thr_max = percentile_threshold(window_scores_max(model, ds_vl), 99.0) # spikes\n",
"\n",
" # Evaluate on injected test windows (subset if limit_test_batches is used)\n",
" ds_te_spike = make_injected_dataset(ds_te, \"spike\")\n",
" ds_te_drift = make_injected_dataset(ds_te, \"drift\")\n",
"\n",
" m_spike = eval_on_injected(model, ds_te_spike, thr_max, score_mode=\"max\")\n",
" m_drift = eval_on_injected(model, ds_te_drift, thr_mean, score_mode=\"mean\")\n",
"\n",
" # Within-window drift detection delay\n",
" if compute_delay:\n",
" ts_thr = timestep_threshold_from_val(model, ds_vl, q=99.0)\n",
" delay = detection_delay_stats(model, ds_te_drift, ts_thr)\n",
" else:\n",
" delay = {\"n\": 0, \"mean\": np.nan, \"median\": np.nan, \"p90\": np.nan}\n",
"\n",
" return {\n",
" \"window\": window,\n",
" \"model\": model_type,\n",
" \"shift_train\": shift_train,\n",
" \"shift_eval\": shift_eval,\n",
" \"train_time_s\": float(t_train),\n",
" \"thr_mean_p99\": float(thr_mean),\n",
" \"thr_max_p99\": float(thr_max),\n",
" \"spike_ap\": float(m_spike[\"ap\"]),\n",
" \"spike_f1\": float(m_spike[\"f1\"]),\n",
" \"spike_fp\": int(m_spike[\"cm\"][0, 1]),\n",
" \"drift_ap\": float(m_drift[\"ap\"]),\n",
" \"drift_f1\": float(m_drift[\"f1\"]),\n",
" \"drift_fp\": int(m_drift[\"cm\"][0, 1]),\n",
" \"drift_delay_median\": float(delay[\"median\"]),\n",
" \"drift_delay_mean\": float(delay[\"mean\"]),\n",
" \"drift_delay_p90\": float(delay[\"p90\"]),\n",
" }\n"
]
},
{
"cell_type": "markdown",
"id": "96dfe809",
"metadata": {},
"source": [
"### 7.1.1 Window length check\n",
"\n",
"The main experiment uses a 24‑hour window (`L=144`) and a larger training stride (`SHIFT_TRAIN=6`) to reduce redundancy.\n",
"To justify these settings in a CPU‑safe way, the notebook runs two compact checks using **CNN only**, **10 epochs**, and limited validation/test batches.\n",
"\n",
"- **Window check:** compare `L ∈ {60, 144, 288}` at a fixed training stride.\n",
"- **Stride check:** compare `SHIFT_train ∈ {1, 6}` at `L=144`.\n",
"\n",
"These checks are not a full hyperparameter search; they provide concrete evidence for the chosen design while keeping runtime reasonable.\n"
]
},
{
"cell_type": "code",
"execution_count": 19,
"id": "8d3edd82",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Epoch 1/10\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:05:33.341272: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"/Library/Frameworks/Python.framework/Versions/3.12/lib/python3.12/site-packages/keras/src/trainers/epoch_iterator.py:164: UserWarning: Your input ran out of data; interrupting training. Make sure that your dataset or generator can generate at least `steps_per_epoch * epochs` batches. You may need to use the `.repeat()` function when building your dataset.\n",
" self._interrupted_warning()\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"96/96 - 3s - 32ms/step - loss: 0.5239 - val_loss: 0.2612 - learning_rate: 0.0010\n",
"Epoch 2/10\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:05:34.104021: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:05:35.886775: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"96/96 - 2s - 25ms/step - loss: 0.2453 - val_loss: 0.2008 - learning_rate: 0.0010\n",
"Epoch 3/10\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:05:36.529052: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:05:38.306397: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"96/96 - 2s - 25ms/step - loss: 0.2025 - val_loss: 0.1684 - learning_rate: 0.0010\n",
"Epoch 4/10\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:05:38.958511: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:05:40.747310: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"96/96 - 2s - 25ms/step - loss: 0.1721 - val_loss: 0.1528 - learning_rate: 0.0010\n",
"Epoch 5/10\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:05:41.380040: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:05:43.183727: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"96/96 - 2s - 25ms/step - loss: 0.1597 - val_loss: 0.1463 - learning_rate: 0.0010\n",
"Epoch 6/10\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:05:43.817673: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:05:45.642877: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"96/96 - 2s - 25ms/step - loss: 0.1526 - val_loss: 0.1392 - learning_rate: 0.0010\n",
"Epoch 7/10\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:05:46.262758: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:05:48.122048: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"96/96 - 2s - 26ms/step - loss: 0.1451 - val_loss: 0.1390 - learning_rate: 0.0010\n",
"Epoch 8/10\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:05:48.742820: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:05:50.633361: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"96/96 - 3s - 26ms/step - loss: 0.1352 - val_loss: 0.1278 - learning_rate: 0.0010\n",
"Epoch 9/10\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:05:51.260270: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:05:53.151249: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"96/96 - 3s - 26ms/step - loss: 0.1279 - val_loss: 0.1224 - learning_rate: 0.0010\n",
"Epoch 10/10\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:05:53.778360: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:05:55.591081: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"96/96 - 2s - 25ms/step - loss: 0.1233 - val_loss: 0.1181 - learning_rate: 0.0010\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:05:56.224488: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:05:56.861606: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"2026-03-17 15:05:57.494376: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"2026-03-17 15:05:58.310325: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"2026-03-17 15:05:59.062005: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"2026-03-17 15:05:59.701978: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"Epoch 1/10\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:06:00.439454: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"2026-03-17 15:06:04.128276: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"/Library/Frameworks/Python.framework/Versions/3.12/lib/python3.12/site-packages/keras/src/trainers/epoch_iterator.py:164: UserWarning: Your input ran out of data; interrupting training. Make sure that your dataset or generator can generate at least `steps_per_epoch * epochs` batches. You may need to use the `.repeat()` function when building your dataset.\n",
" self._interrupted_warning()\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"96/96 - 5s - 49ms/step - loss: 0.5087 - val_loss: 0.2832 - learning_rate: 0.0010\n",
"Epoch 2/10\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:06:05.185824: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:06:08.126577: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"96/96 - 4s - 41ms/step - loss: 0.2589 - val_loss: 0.2164 - learning_rate: 0.0010\n",
"Epoch 3/10\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:06:09.125445: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:06:12.099395: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"96/96 - 4s - 41ms/step - loss: 0.2269 - val_loss: 0.1959 - learning_rate: 0.0010\n",
"Epoch 4/10\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:06:13.082423: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:06:15.999814: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"96/96 - 4s - 41ms/step - loss: 0.2034 - val_loss: 0.1791 - learning_rate: 0.0010\n",
"Epoch 5/10\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:06:16.984788: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:06:19.882235: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"96/96 - 4s - 41ms/step - loss: 0.1816 - val_loss: 0.1661 - learning_rate: 0.0010\n",
"Epoch 6/10\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:06:20.921709: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:06:23.900853: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"96/96 - 4s - 41ms/step - loss: 0.1709 - val_loss: 0.1582 - learning_rate: 0.0010\n",
"Epoch 7/10\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:06:24.889156: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:06:27.789001: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"96/96 - 4s - 41ms/step - loss: 0.1642 - val_loss: 0.1539 - learning_rate: 0.0010\n",
"Epoch 8/10\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:06:28.784545: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:06:31.801481: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"96/96 - 4s - 42ms/step - loss: 0.1560 - val_loss: 0.1471 - learning_rate: 0.0010\n",
"Epoch 9/10\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:06:32.785344: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:06:35.771278: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"96/96 - 4s - 42ms/step - loss: 0.1508 - val_loss: 0.1434 - learning_rate: 0.0010\n",
"Epoch 10/10\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:06:36.777844: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:06:39.807344: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"96/96 - 4s - 42ms/step - loss: 0.1475 - val_loss: 0.1414 - learning_rate: 0.0010\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:06:40.819369: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:06:41.874969: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"2026-03-17 15:06:42.926886: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"2026-03-17 15:06:44.121508: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"2026-03-17 15:06:45.297487: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"2026-03-17 15:06:46.357975: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"Epoch 1/10\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:06:47.532694: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"2026-03-17 15:06:52.926378: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"/Library/Frameworks/Python.framework/Versions/3.12/lib/python3.12/site-packages/keras/src/trainers/epoch_iterator.py:164: UserWarning: Your input ran out of data; interrupting training. Make sure that your dataset or generator can generate at least `steps_per_epoch * epochs` batches. You may need to use the `.repeat()` function when building your dataset.\n",
" self._interrupted_warning()\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"96/96 - 7s - 73ms/step - loss: 0.4493 - val_loss: 0.2755 - learning_rate: 0.0010\n",
"Epoch 2/10\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:06:54.575935: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:06:59.080053: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"96/96 - 6s - 63ms/step - loss: 0.2665 - val_loss: 0.2255 - learning_rate: 0.0010\n",
"Epoch 3/10\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:07:00.647208: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:07:05.346018: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"96/96 - 6s - 66ms/step - loss: 0.2293 - val_loss: 0.1979 - learning_rate: 0.0010\n",
"Epoch 4/10\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:07:07.004067: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:07:11.763797: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"96/96 - 6s - 66ms/step - loss: 0.2048 - val_loss: 0.1811 - learning_rate: 0.0010\n",
"Epoch 5/10\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:07:13.373432: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:07:17.995147: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"96/96 - 6s - 65ms/step - loss: 0.1915 - val_loss: 0.1756 - learning_rate: 0.0010\n",
"Epoch 6/10\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:07:19.585339: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:07:24.204778: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"96/96 - 6s - 65ms/step - loss: 0.1820 - val_loss: 0.1656 - learning_rate: 0.0010\n",
"Epoch 7/10\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:07:25.810630: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:07:30.494381: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"96/96 - 6s - 66ms/step - loss: 0.1730 - val_loss: 0.1615 - learning_rate: 0.0010\n",
"Epoch 8/10\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:07:32.156716: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:07:36.754877: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"96/96 - 6s - 64ms/step - loss: 0.1671 - val_loss: 0.1562 - learning_rate: 0.0010\n",
"Epoch 9/10\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:07:38.318922: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:07:43.082973: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"96/96 - 6s - 67ms/step - loss: 0.1634 - val_loss: 0.1541 - learning_rate: 0.0010\n",
"Epoch 10/10\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:07:44.703114: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:07:49.520465: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"96/96 - 6s - 67ms/step - loss: 0.1604 - val_loss: 0.1543 - learning_rate: 0.0010\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:07:51.163809: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:07:53.009604: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"2026-03-17 15:07:54.737652: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"2026-03-17 15:07:56.550418: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"2026-03-17 15:07:58.343208: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"2026-03-17 15:08:00.093711: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"2026-03-17 15:08:01.906944: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n"
]
},
{
"data": {
"text/html": [
"\n",
"\n",
"
\n",
" \n",
" \n",
" | \n",
" window | \n",
" train_time_s | \n",
" spike_f1 | \n",
" drift_f1 | \n",
" delay_median_hours | \n",
" delay_p90_hours | \n",
" spike_fp | \n",
" drift_fp | \n",
"
\n",
" \n",
" \n",
" \n",
" | 0 | \n",
" 60 | \n",
" 25.171580 | \n",
" 0.991418 | \n",
" 0.875874 | \n",
" 4.333333 | \n",
" 8.333333 | \n",
" 55 | \n",
" 2 | \n",
"
\n",
" \n",
" | 1 | \n",
" 144 | \n",
" 40.344092 | \n",
" 0.994620 | \n",
" 0.881503 | \n",
" 8.833333 | \n",
" 18.333333 | \n",
" 0 | \n",
" 1 | \n",
"
\n",
" \n",
" | 2 | \n",
" 288 | \n",
" 63.594538 | \n",
" 0.992867 | \n",
" 0.922923 | \n",
" 16.666667 | \n",
" 36.000000 | \n",
" 0 | \n",
" 15 | \n",
"
\n",
" \n",
"
\n",
"
\n",
"\n",
"
\n",
" \n",
" \n",
" | \n",
" shift_train | \n",
" train_time_s | \n",
" spike_f1 | \n",
" drift_f1 | \n",
" delay_median_hours | \n",
" delay_p90_hours | \n",
" spike_fp | \n",
" drift_fp | \n",
"
\n",
" \n",
" \n",
" \n",
" | 0 | \n",
" 1 | \n",
" 320.026314 | \n",
" 0.996526 | \n",
" 0.902154 | \n",
" 8.833333 | \n",
" 19.0 | \n",
" 0 | \n",
" 24 | \n",
"
\n",
" \n",
" | 1 | \n",
" 6 | \n",
" 65.706250 | \n",
" 0.984335 | \n",
" 0.908813 | \n",
" 8.333333 | \n",
" 18.5 | \n",
" 0 | \n",
" 0 | \n",
"
\n",
" \n",
"
\n",
"
Model: \"cnn_autoencoder\"\n",
"\n"
],
"text/plain": [
"\u001b[1mModel: \"cnn_autoencoder\"\u001b[0m\n"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"text/html": [
"┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓\n",
"┃ Layer (type) ┃ Output Shape ┃ Param # ┃\n",
"┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩\n",
"│ input_layer_5 (InputLayer) │ (None, 144, 14) │ 0 │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ conv1d_20 (Conv1D) │ (None, 144, 32) │ 1,376 │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ max_pooling1d_10 (MaxPooling1D) │ (None, 72, 32) │ 0 │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ conv1d_21 (Conv1D) │ (None, 72, 64) │ 6,208 │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ max_pooling1d_11 (MaxPooling1D) │ (None, 36, 64) │ 0 │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ flatten_5 (Flatten) │ (None, 2304) │ 0 │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ dense_10 (Dense) │ (None, 32) │ 73,760 │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ dense_11 (Dense) │ (None, 2304) │ 76,032 │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ reshape_5 (Reshape) │ (None, 36, 64) │ 0 │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ up_sampling1d_10 (UpSampling1D) │ (None, 72, 64) │ 0 │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ conv1d_22 (Conv1D) │ (None, 72, 64) │ 12,352 │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ up_sampling1d_11 (UpSampling1D) │ (None, 144, 64) │ 0 │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ conv1d_23 (Conv1D) │ (None, 144, 14) │ 2,702 │\n",
"└─────────────────────────────────┴────────────────────────┴───────────────┘\n",
"\n"
],
"text/plain": [
"┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓\n",
"┃\u001b[1m \u001b[0m\u001b[1mLayer (type) \u001b[0m\u001b[1m \u001b[0m┃\u001b[1m \u001b[0m\u001b[1mOutput Shape \u001b[0m\u001b[1m \u001b[0m┃\u001b[1m \u001b[0m\u001b[1m Param #\u001b[0m\u001b[1m \u001b[0m┃\n",
"┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩\n",
"│ input_layer_5 (\u001b[38;5;33mInputLayer\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m144\u001b[0m, \u001b[38;5;34m14\u001b[0m) │ \u001b[38;5;34m0\u001b[0m │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ conv1d_20 (\u001b[38;5;33mConv1D\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m144\u001b[0m, \u001b[38;5;34m32\u001b[0m) │ \u001b[38;5;34m1,376\u001b[0m │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ max_pooling1d_10 (\u001b[38;5;33mMaxPooling1D\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m72\u001b[0m, \u001b[38;5;34m32\u001b[0m) │ \u001b[38;5;34m0\u001b[0m │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ conv1d_21 (\u001b[38;5;33mConv1D\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m72\u001b[0m, \u001b[38;5;34m64\u001b[0m) │ \u001b[38;5;34m6,208\u001b[0m │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ max_pooling1d_11 (\u001b[38;5;33mMaxPooling1D\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m36\u001b[0m, \u001b[38;5;34m64\u001b[0m) │ \u001b[38;5;34m0\u001b[0m │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ flatten_5 (\u001b[38;5;33mFlatten\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m2304\u001b[0m) │ \u001b[38;5;34m0\u001b[0m │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ dense_10 (\u001b[38;5;33mDense\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m32\u001b[0m) │ \u001b[38;5;34m73,760\u001b[0m │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ dense_11 (\u001b[38;5;33mDense\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m2304\u001b[0m) │ \u001b[38;5;34m76,032\u001b[0m │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ reshape_5 (\u001b[38;5;33mReshape\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m36\u001b[0m, \u001b[38;5;34m64\u001b[0m) │ \u001b[38;5;34m0\u001b[0m │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ up_sampling1d_10 (\u001b[38;5;33mUpSampling1D\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m72\u001b[0m, \u001b[38;5;34m64\u001b[0m) │ \u001b[38;5;34m0\u001b[0m │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ conv1d_22 (\u001b[38;5;33mConv1D\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m72\u001b[0m, \u001b[38;5;34m64\u001b[0m) │ \u001b[38;5;34m12,352\u001b[0m │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ up_sampling1d_11 (\u001b[38;5;33mUpSampling1D\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m144\u001b[0m, \u001b[38;5;34m64\u001b[0m) │ \u001b[38;5;34m0\u001b[0m │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ conv1d_23 (\u001b[38;5;33mConv1D\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m144\u001b[0m, \u001b[38;5;34m14\u001b[0m) │ \u001b[38;5;34m2,702\u001b[0m │\n",
"└─────────────────────────────────┴────────────────────────┴───────────────┘\n"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"text/html": [
" Total params: 172,430 (673.55 KB)\n",
"\n"
],
"text/plain": [
"\u001b[1m Total params: \u001b[0m\u001b[38;5;34m172,430\u001b[0m (673.55 KB)\n"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"text/html": [
" Trainable params: 172,430 (673.55 KB)\n",
"\n"
],
"text/plain": [
"\u001b[1m Trainable params: \u001b[0m\u001b[38;5;34m172,430\u001b[0m (673.55 KB)\n"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"text/html": [
" Non-trainable params: 0 (0.00 B)\n",
"\n"
],
"text/plain": [
"\u001b[1m Non-trainable params: \u001b[0m\u001b[38;5;34m0\u001b[0m (0.00 B)\n"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"text/html": [
"Model: \"lstm_autoencoder\"\n",
"\n"
],
"text/plain": [
"\u001b[1mModel: \"lstm_autoencoder\"\u001b[0m\n"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"text/html": [
"┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓\n",
"┃ Layer (type) ┃ Output Shape ┃ Param # ┃\n",
"┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩\n",
"│ input_layer_6 (InputLayer) │ (None, 144, 14) │ 0 │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ lstm (LSTM) │ (None, 32) │ 6,016 │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ repeat_vector (RepeatVector) │ (None, 144, 32) │ 0 │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ lstm_1 (LSTM) │ (None, 144, 32) │ 8,320 │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ time_distributed │ (None, 144, 14) │ 462 │\n",
"│ (TimeDistributed) │ │ │\n",
"└─────────────────────────────────┴────────────────────────┴───────────────┘\n",
"\n"
],
"text/plain": [
"┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓\n",
"┃\u001b[1m \u001b[0m\u001b[1mLayer (type) \u001b[0m\u001b[1m \u001b[0m┃\u001b[1m \u001b[0m\u001b[1mOutput Shape \u001b[0m\u001b[1m \u001b[0m┃\u001b[1m \u001b[0m\u001b[1m Param #\u001b[0m\u001b[1m \u001b[0m┃\n",
"┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩\n",
"│ input_layer_6 (\u001b[38;5;33mInputLayer\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m144\u001b[0m, \u001b[38;5;34m14\u001b[0m) │ \u001b[38;5;34m0\u001b[0m │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ lstm (\u001b[38;5;33mLSTM\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m32\u001b[0m) │ \u001b[38;5;34m6,016\u001b[0m │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ repeat_vector (\u001b[38;5;33mRepeatVector\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m144\u001b[0m, \u001b[38;5;34m32\u001b[0m) │ \u001b[38;5;34m0\u001b[0m │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ lstm_1 (\u001b[38;5;33mLSTM\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m144\u001b[0m, \u001b[38;5;34m32\u001b[0m) │ \u001b[38;5;34m8,320\u001b[0m │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ time_distributed │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m144\u001b[0m, \u001b[38;5;34m14\u001b[0m) │ \u001b[38;5;34m462\u001b[0m │\n",
"│ (\u001b[38;5;33mTimeDistributed\u001b[0m) │ │ │\n",
"└─────────────────────────────────┴────────────────────────┴───────────────┘\n"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"text/html": [
" Total params: 14,798 (57.80 KB)\n",
"\n"
],
"text/plain": [
"\u001b[1m Total params: \u001b[0m\u001b[38;5;34m14,798\u001b[0m (57.80 KB)\n"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"text/html": [
" Trainable params: 14,798 (57.80 KB)\n",
"\n"
],
"text/plain": [
"\u001b[1m Trainable params: \u001b[0m\u001b[38;5;34m14,798\u001b[0m (57.80 KB)\n"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"text/html": [
" Non-trainable params: 0 (0.00 B)\n",
"\n"
],
"text/plain": [
"\u001b[1m Non-trainable params: \u001b[0m\u001b[38;5;34m0\u001b[0m (0.00 B)\n"
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"# Build models\n",
"cnn_ae = build_cnn_autoencoder(WINDOW, F, latent=32)\n",
"lstm_ae = build_lstm_autoencoder(WINDOW, F, latent=32)\n",
"\n",
"cnn_ae.summary()\n",
"lstm_ae.summary()\n"
]
},
{
"cell_type": "code",
"execution_count": 22,
"id": "ec1072db",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Epoch 1/30\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:14:47.080768: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"/Library/Frameworks/Python.framework/Versions/3.12/lib/python3.12/site-packages/keras/src/trainers/epoch_iterator.py:164: UserWarning: Your input ran out of data; interrupting training. Make sure that your dataset or generator can generate at least `steps_per_epoch * epochs` batches. You may need to use the `.repeat()` function when building your dataset.\n",
" self._interrupted_warning()\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"192/192 - 11s - 56ms/step - loss: 0.3512 - val_loss: 0.7175 - learning_rate: 0.0010\n",
"Epoch 2/30\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:14:51.921003: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:14:57.820547: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"192/192 - 11s - 58ms/step - loss: 0.1824 - val_loss: 0.6494 - learning_rate: 0.0010\n",
"Epoch 3/30\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:15:02.986598: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:15:08.422343: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"192/192 - 10s - 55ms/step - loss: 0.1527 - val_loss: 0.6202 - learning_rate: 0.0010\n",
"Epoch 4/30\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:15:13.462349: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:15:19.707329: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"192/192 - 11s - 59ms/step - loss: 0.1396 - val_loss: 0.6034 - learning_rate: 0.0010\n",
"Epoch 5/30\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:15:24.814067: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:15:30.701646: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"192/192 - 11s - 57ms/step - loss: 0.1322 - val_loss: 0.5983 - learning_rate: 0.0010\n",
"Epoch 6/30\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:15:35.712849: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:15:41.712864: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"192/192 - 11s - 57ms/step - loss: 0.1279 - val_loss: 0.5897 - learning_rate: 0.0010\n",
"Epoch 7/30\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:15:46.604879: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:15:52.552082: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"192/192 - 11s - 57ms/step - loss: 0.1245 - val_loss: 0.5782 - learning_rate: 0.0010\n",
"Epoch 8/30\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:15:57.522366: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:16:03.474493: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"192/192 - 11s - 57ms/step - loss: 0.1224 - val_loss: 0.5715 - learning_rate: 0.0010\n",
"Epoch 9/30\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:16:08.427862: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:16:15.567421: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"192/192 - 12s - 63ms/step - loss: 0.1199 - val_loss: 0.5675 - learning_rate: 0.0010\n",
"Epoch 10/30\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:16:20.477376: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:16:26.241189: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"192/192 - 11s - 56ms/step - loss: 0.1185 - val_loss: 0.5704 - learning_rate: 0.0010\n",
"Epoch 11/30\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:16:31.170491: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:16:36.904349: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"192/192 - 11s - 55ms/step - loss: 0.1172 - val_loss: 0.5688 - learning_rate: 0.0010\n",
"Epoch 12/30\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:16:41.775748: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:16:47.361579: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"192/192 - 10s - 55ms/step - loss: 0.1124 - val_loss: 0.5579 - learning_rate: 5.0000e-04\n",
"Epoch 13/30\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:16:52.250995: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:16:57.845362: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"192/192 - 10s - 54ms/step - loss: 0.1118 - val_loss: 0.5569 - learning_rate: 5.0000e-04\n",
"Epoch 14/30\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:17:02.713483: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:17:08.180810: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"192/192 - 10s - 54ms/step - loss: 0.1114 - val_loss: 0.5576 - learning_rate: 5.0000e-04\n",
"Epoch 15/30\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:17:12.994188: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:17:18.534077: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"192/192 - 10s - 54ms/step - loss: 0.1112 - val_loss: 0.5584 - learning_rate: 5.0000e-04\n",
"Epoch 16/30\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:17:23.336839: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:17:28.968049: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"192/192 - 10s - 54ms/step - loss: 0.1100 - val_loss: 0.5566 - learning_rate: 2.5000e-04\n",
"Epoch 17/30\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:17:33.731054: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:17:39.231012: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"192/192 - 10s - 53ms/step - loss: 0.1098 - val_loss: 0.5571 - learning_rate: 2.5000e-04\n",
"Epoch 18/30\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:17:43.985866: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:17:49.405451: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"192/192 - 10s - 53ms/step - loss: 0.1096 - val_loss: 0.5581 - learning_rate: 2.5000e-04\n",
"Epoch 19/30\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:17:54.212578: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:17:59.798352: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"192/192 - 11s - 55ms/step - loss: 0.1092 - val_loss: 0.5569 - learning_rate: 1.2500e-04\n",
"Epoch 20/30\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:18:04.813409: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:18:10.594743: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"192/192 - 11s - 56ms/step - loss: 0.1090 - val_loss: 0.5570 - learning_rate: 1.2500e-04\n",
"Epoch 21/30\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:18:15.619909: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:18:21.227679: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"192/192 - 11s - 55ms/step - loss: 0.1088 - val_loss: 0.5568 - learning_rate: 6.2500e-05\n",
"Epoch 1/30\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:18:26.132601: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:19:01.038489: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"192/192 - 46s - 242ms/step - loss: 0.4851 - val_loss: 0.6609 - learning_rate: 0.0010\n",
"Epoch 2/30\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:19:12.525418: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:19:45.193998: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"192/192 - 43s - 224ms/step - loss: 0.3940 - val_loss: 0.6204 - learning_rate: 0.0010\n",
"Epoch 3/30\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:19:55.565536: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:20:26.357000: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"192/192 - 41s - 215ms/step - loss: 0.3663 - val_loss: 0.6023 - learning_rate: 0.0010\n",
"Epoch 4/30\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:20:36.792086: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:21:09.354480: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"192/192 - 43s - 227ms/step - loss: 0.3517 - val_loss: 0.5917 - learning_rate: 0.0010\n",
"Epoch 5/30\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:21:20.286880: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:21:52.280170: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"192/192 - 42s - 220ms/step - loss: 0.3448 - val_loss: 0.5858 - learning_rate: 0.0010\n",
"Epoch 6/30\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:22:02.477657: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:22:34.778628: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"192/192 - 42s - 220ms/step - loss: 0.3402 - val_loss: 0.5819 - learning_rate: 0.0010\n",
"Epoch 7/30\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:22:44.724432: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:23:16.236082: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"192/192 - 42s - 216ms/step - loss: 0.3365 - val_loss: 0.5781 - learning_rate: 0.0010\n",
"Epoch 8/30\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:23:26.257193: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:23:58.053334: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"192/192 - 42s - 220ms/step - loss: 0.3271 - val_loss: 0.5614 - learning_rate: 0.0010\n",
"Epoch 9/30\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:24:08.436830: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:24:38.830651: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"192/192 - 41s - 211ms/step - loss: 0.2835 - val_loss: 0.5089 - learning_rate: 0.0010\n",
"Epoch 10/30\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:24:49.033916: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:25:20.667770: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"192/192 - 42s - 218ms/step - loss: 0.2569 - val_loss: 0.4868 - learning_rate: 0.0010\n",
"Epoch 11/30\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:25:30.823961: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:26:02.548977: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"192/192 - 42s - 217ms/step - loss: 0.2438 - val_loss: 0.4793 - learning_rate: 0.0010\n",
"Epoch 12/30\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:26:12.521423: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:26:44.363131: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"192/192 - 42s - 218ms/step - loss: 0.2362 - val_loss: 0.4718 - learning_rate: 0.0010\n",
"Epoch 13/30\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:26:54.351983: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:27:25.683872: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"192/192 - 42s - 217ms/step - loss: 0.2309 - val_loss: 0.4688 - learning_rate: 0.0010\n",
"Epoch 14/30\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:27:35.995323: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:28:07.115190: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"192/192 - 41s - 213ms/step - loss: 0.2266 - val_loss: 0.4643 - learning_rate: 0.0010\n",
"Epoch 15/30\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:28:16.843894: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:28:47.935345: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"192/192 - 41s - 214ms/step - loss: 0.2234 - val_loss: 0.4607 - learning_rate: 0.0010\n",
"Epoch 16/30\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:28:57.939601: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:29:27.595484: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"192/192 - 40s - 209ms/step - loss: 0.2208 - val_loss: 0.4582 - learning_rate: 0.0010\n",
"Epoch 17/30\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:29:38.083653: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:30:09.284464: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"192/192 - 41s - 216ms/step - loss: 0.2185 - val_loss: 0.4560 - learning_rate: 0.0010\n",
"Epoch 18/30\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:30:19.569921: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:30:50.948949: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"192/192 - 42s - 218ms/step - loss: 0.2166 - val_loss: 0.4543 - learning_rate: 0.0010\n",
"Epoch 19/30\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:31:01.419102: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:31:33.129528: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"192/192 - 42s - 218ms/step - loss: 0.2150 - val_loss: 0.4529 - learning_rate: 0.0010\n",
"Epoch 20/30\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:31:43.348756: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:32:15.284683: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"192/192 - 42s - 218ms/step - loss: 0.2135 - val_loss: 0.4518 - learning_rate: 0.0010\n",
"Epoch 21/30\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:32:25.127627: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:32:54.883262: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"192/192 - 40s - 206ms/step - loss: 0.2123 - val_loss: 0.4509 - learning_rate: 0.0010\n",
"Epoch 22/30\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:33:04.683845: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:33:35.114766: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"192/192 - 41s - 213ms/step - loss: 0.2111 - val_loss: 0.4494 - learning_rate: 0.0010\n",
"Epoch 23/30\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:33:45.550932: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:34:15.671978: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"192/192 - 40s - 209ms/step - loss: 0.2102 - val_loss: 0.4484 - learning_rate: 0.0010\n",
"Epoch 24/30\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:34:25.655032: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:34:55.756952: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"192/192 - 40s - 211ms/step - loss: 0.2091 - val_loss: 0.4478 - learning_rate: 0.0010\n",
"Epoch 25/30\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:35:06.077536: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:35:37.807001: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"192/192 - 42s - 219ms/step - loss: 0.2082 - val_loss: 0.4467 - learning_rate: 0.0010\n",
"Epoch 26/30\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:35:48.157176: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:36:23.368549: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"192/192 - 46s - 239ms/step - loss: 0.2071 - val_loss: 0.4446 - learning_rate: 0.0010\n",
"Epoch 27/30\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:36:34.047099: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:37:04.980969: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"192/192 - 41s - 215ms/step - loss: 0.2059 - val_loss: 0.4435 - learning_rate: 0.0010\n",
"Epoch 28/30\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:37:15.334376: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:37:46.408717: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"192/192 - 41s - 214ms/step - loss: 0.2041 - val_loss: 0.4421 - learning_rate: 0.0010\n",
"Epoch 29/30\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:37:56.416870: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:38:28.361354: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"192/192 - 42s - 216ms/step - loss: 0.2021 - val_loss: 0.4391 - learning_rate: 0.0010\n",
"Epoch 30/30\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:38:37.967796: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n",
"2026-03-17 15:39:08.991314: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"192/192 - 41s - 213ms/step - loss: 0.1992 - val_loss: 0.4360 - learning_rate: 0.0010\n",
"CNN training time (s): 225.0\n",
"LSTM training time (s): 1252.8\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:39:18.947102: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"\t [[{{node IteratorGetNext}}]]\n"
]
}
],
"source": [
"# Main training run (denoising AE: (x + noise) → x)\n",
"EPOCHS_MAIN = 30\n",
"\n",
"cnn_hist, cnn_time = train_autoencoder(cnn_ae, ds_train_fit, ds_val_fit, epochs=EPOCHS_MAIN)\n",
"lstm_hist, lstm_time = train_autoencoder(lstm_ae, ds_train_fit, ds_val_fit, epochs=EPOCHS_MAIN)\n",
"\n",
"print(f\"CNN training time (s): {cnn_time:.1f}\")\n",
"print(f\"LSTM training time (s): {lstm_time:.1f}\")"
]
},
{
"cell_type": "code",
"execution_count": 23,
"id": "8cf026db",
"metadata": {},
"outputs": [
{
"data": {
"image/png": "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",
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"image/png": "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",
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"def plot_history(history, title: str):\n",
" fig, ax = plt.subplots(figsize=(8, 3))\n",
" ax.plot(history.history[\"loss\"], label=\"train\")\n",
" ax.plot(history.history[\"val_loss\"], label=\"val\")\n",
" ax.set_title(title)\n",
" ax.set_xlabel(\"epoch\")\n",
" ax.set_ylabel(\"MAE loss\")\n",
" ax.legend()\n",
" plt.show()\n",
"\n",
"plot_history(cnn_hist, \"CNN Autoencoder training curve\")\n",
"plot_history(lstm_hist, \"LSTM Autoencoder training curve\")\n"
]
},
{
"cell_type": "markdown",
"id": "ab57f3f9",
"metadata": {},
"source": [
"## 9. Reconstruction error and thresholding\n",
"\n",
"Each window receives an anomaly score computed from reconstruction error:\n",
"\n",
"- **mean score:** average absolute error over all timesteps/features (drift‑sensitive)\n",
"- **max score:** maximum absolute error in the window (spike‑sensitive)\n",
"\n",
"Thresholds are set from **normal validation** scores using the 99th percentile (p99), which fixes the expected false‑alarm rate on normal data.\n"
]
},
{
"cell_type": "code",
"execution_count": 24,
"id": "e6003dc4",
"metadata": {},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:39:24.179282: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"2026-03-17 15:40:22.208763: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"2026-03-17 15:40:27.288296: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"CNN threshold mean (p99): 0.19743414118886002\n",
"LSTM threshold mean (p99): 0.40897981226444247\n",
"CNN threshold max (p99): 4.13346616268158\n",
"LSTM threshold max (p99): 5.07332712173462\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:41:23.951428: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n"
]
}
],
"source": [
"# Thresholds calibrated on clean validation windows (p99)\n",
"# Mean score -> drift; Max score -> spikes.\n",
"\n",
"val_scores_mean_cnn = window_scores_mean(cnn_ae, ds_val)\n",
"val_scores_mean_lstm = window_scores_mean(lstm_ae, ds_val)\n",
"\n",
"thr_mean_cnn = percentile_threshold(val_scores_mean_cnn, 99.0)\n",
"thr_mean_lstm = percentile_threshold(val_scores_mean_lstm, 99.0)\n",
"\n",
"val_scores_max_cnn = window_scores_max(cnn_ae, ds_val)\n",
"val_scores_max_lstm = window_scores_max(lstm_ae, ds_val)\n",
"\n",
"thr_max_cnn = percentile_threshold(val_scores_max_cnn, 99.0)\n",
"thr_max_lstm = percentile_threshold(val_scores_max_lstm, 99.0)\n",
"\n",
"print(\"CNN threshold mean (p99):\", thr_mean_cnn)\n",
"print(\"LSTM threshold mean (p99):\", thr_mean_lstm)\n",
"print(\"CNN threshold max (p99):\", thr_max_cnn)\n",
"print(\"LSTM threshold max (p99):\", thr_max_lstm)\n"
]
},
{
"cell_type": "markdown",
"id": "5b55bc1d",
"metadata": {},
"source": [
"## 10. Evaluation on injected anomalies (window-level)\n",
"\n",
"Evaluation uses injected anomalies with ground-truth labels.\n",
"Metrics:\n",
"- PR‑AUC (Average Precision)\n",
"- Precision, Recall, F1 at the chosen threshold\n",
"- Confusion matrix"
]
},
{
"cell_type": "code",
"execution_count": 25,
"id": "6243c2a8",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"=== CNN - spike (max score) ===\n",
"AP (PR-AUC): 0.9999420942684678\n",
"Precision: 0.9998094149037545\n",
"Recall: 0.9990478004189678\n",
"F1: 0.999428462564298\n",
"Confusion matrix [[TN, FP],[FN, TP]]:\n",
"[[47184 3]\n",
" [ 15 15738]]\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:41:29.426275: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n"
]
},
{
"data": {
"image/png": "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",
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n",
"=== LSTM - spike (max score) ===\n",
"AP (PR-AUC): 0.9973633106096367\n",
"Precision: 0.9948123478929166\n",
"Recall: 0.9860344061448613\n",
"F1: 0.990403927694711\n",
"Confusion matrix [[TN, FP],[FN, TP]]:\n",
"[[47106 81]\n",
" [ 220 15533]]\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:42:25.488616: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n"
]
},
{
"data": {
"image/png": "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",
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n",
"=== CNN - drift (mean score) ===\n",
"AP (PR-AUC): 0.9749031915981917\n",
"Precision: 0.9944611875865439\n",
"Recall: 0.7741424132902163\n",
"F1: 0.8705790074158585\n",
"Confusion matrix [[TN, FP],[FN, TP]]:\n",
"[[47101 68]\n",
" [ 3562 12209]]\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:42:31.024979: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n"
]
},
{
"data": {
"image/png": "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",
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n",
"=== LSTM - drift (mean score) ===\n",
"AP (PR-AUC): 0.9552245181677891\n",
"Precision: 0.9970822964043594\n",
"Recall: 0.7367319764124025\n",
"F1: 0.8473599766627772\n",
"Confusion matrix [[TN, FP],[FN, TP]]:\n",
"[[47135 34]\n",
" [ 4152 11619]]\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:43:27.351240: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n"
]
},
{
"data": {
"image/png": "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",
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"# Window-level evaluation on injected anomalies\n",
"# Spikes: max score; Drift: mean score.\n",
"\n",
"print(\"=== CNN - spike (max score) ===\")\n",
"m_cnn_spike = eval_on_injected(cnn_ae, make_injected_dataset(ds_test, \"spike\"), thr_max_cnn, score_mode=\"max\")\n",
"print(\"AP (PR-AUC):\", m_cnn_spike[\"ap\"])\n",
"print(\"Precision:\", m_cnn_spike[\"precision\"])\n",
"print(\"Recall:\", m_cnn_spike[\"recall\"])\n",
"print(\"F1:\", m_cnn_spike[\"f1\"])\n",
"print(\"Confusion matrix [[TN, FP],[FN, TP]]:\")\n",
"print(m_cnn_spike[\"cm\"])\n",
"plot_confusion_matrix(m_cnn_spike[\"cm\"], \"CNN — Spike (max score)\")\n",
"\n",
"print(\"\\n=== LSTM - spike (max score) ===\")\n",
"m_lstm_spike = eval_on_injected(lstm_ae, make_injected_dataset(ds_test, \"spike\"), thr_max_lstm, score_mode=\"max\")\n",
"print(\"AP (PR-AUC):\", m_lstm_spike[\"ap\"])\n",
"print(\"Precision:\", m_lstm_spike[\"precision\"])\n",
"print(\"Recall:\", m_lstm_spike[\"recall\"])\n",
"print(\"F1:\", m_lstm_spike[\"f1\"])\n",
"print(\"Confusion matrix [[TN, FP],[FN, TP]]:\")\n",
"print(m_lstm_spike[\"cm\"])\n",
"plot_confusion_matrix(m_lstm_spike[\"cm\"], \"LSTM — Spike (max score)\")\n",
"\n",
"print(\"\\n=== CNN - drift (mean score) ===\")\n",
"m_cnn_drift = eval_on_injected(cnn_ae, make_injected_dataset(ds_test, \"drift\"), thr_mean_cnn, score_mode=\"mean\")\n",
"print(\"AP (PR-AUC):\", m_cnn_drift[\"ap\"])\n",
"print(\"Precision:\", m_cnn_drift[\"precision\"])\n",
"print(\"Recall:\", m_cnn_drift[\"recall\"])\n",
"print(\"F1:\", m_cnn_drift[\"f1\"])\n",
"print(\"Confusion matrix [[TN, FP],[FN, TP]]:\")\n",
"print(m_cnn_drift[\"cm\"])\n",
"plot_confusion_matrix(m_cnn_drift[\"cm\"], \"CNN — Drift (mean score)\")\n",
"\n",
"print(\"\\n=== LSTM - drift (mean score) ===\")\n",
"m_lstm_drift = eval_on_injected(lstm_ae, make_injected_dataset(ds_test, \"drift\"), thr_mean_lstm, score_mode=\"mean\")\n",
"print(\"AP (PR-AUC):\", m_lstm_drift[\"ap\"])\n",
"print(\"Precision:\", m_lstm_drift[\"precision\"])\n",
"print(\"Recall:\", m_lstm_drift[\"recall\"])\n",
"print(\"F1:\", m_lstm_drift[\"f1\"])\n",
"print(\"Confusion matrix [[TN, FP],[FN, TP]]:\")\n",
"print(m_lstm_drift[\"cm\"])\n",
"plot_confusion_matrix(m_lstm_drift[\"cm\"], \"LSTM — Drift (mean score)\")\n"
]
},
{
"cell_type": "markdown",
"id": "508b2d9c",
"metadata": {},
"source": [
"## 11. Detection delay (within‑window)\n",
"\n",
"For drift, delay is measured as the first timestep where per‑timestep reconstruction error exceeds a threshold, minus the drift start index `t0`.\n",
"\n",
"Delay is reported in timesteps and later converted to hours (10‑minute sampling).\n"
]
},
{
"cell_type": "code",
"execution_count": 26,
"id": "09e36d4e",
"metadata": {},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:43:32.401105: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"2026-03-17 15:44:27.440390: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"2026-03-17 15:44:33.025085: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"CNN drift delay stats (timesteps): {'n': 14228, 'mean': 60.628408771436604, 'median': 56.0, 'p90': 115.0}\n",
"LSTM drift delay stats (timesteps): {'n': 13105, 'mean': 10.022663105684853, 'median': 0.0, 'p90': 28.0}\n",
"CNN drift delay (hours): {'median_h': 9.333333333333332, 'mean_h': 10.104734795239434, 'p90_h': 19.166666666666664}\n",
"LSTM drift delay (hours): {'median_h': 0.0, 'mean_h': 1.6704438509474753, 'p90_h': 4.666666666666666}\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:45:29.112210: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n"
]
}
],
"source": [
"# Within-window drift detection delay (timesteps, then converted to hours)\n",
"# Threshold is derived from per-timestep scores on clean validation windows (p99).\n",
"\n",
"ts_thr_cnn = timestep_threshold_from_val(cnn_ae, ds_val, q=99.0)\n",
"ts_thr_lstm = timestep_threshold_from_val(lstm_ae, ds_val, q=99.0)\n",
"\n",
"delay_cnn_drift = detection_delay_stats(cnn_ae, make_injected_dataset(ds_test, \"drift\"), ts_thr_cnn)\n",
"delay_lstm_drift = detection_delay_stats(lstm_ae, make_injected_dataset(ds_test, \"drift\"), ts_thr_lstm)\n",
"\n",
"print(\"CNN drift delay stats (timesteps):\", delay_cnn_drift)\n",
"print(\"LSTM drift delay stats (timesteps):\", delay_lstm_drift)\n",
"\n",
"def _delay_hours(d):\n",
" if d[\"n\"] == 0 or np.isnan(d[\"median\"]):\n",
" return {\"median_h\": np.nan, \"mean_h\": np.nan, \"p90_h\": np.nan}\n",
" return {\n",
" \"median_h\": d[\"median\"] * (10.0 / 60.0),\n",
" \"mean_h\": d[\"mean\"] * (10.0 / 60.0),\n",
" \"p90_h\": d[\"p90\"] * (10.0 / 60.0),\n",
" }\n",
"\n",
"print(\"CNN drift delay (hours):\", _delay_hours(delay_cnn_drift))\n",
"print(\"LSTM drift delay (hours):\", _delay_hours(delay_lstm_drift))\n"
]
},
{
"cell_type": "markdown",
"id": "0a456e3a",
"metadata": {},
"source": [
"## 12. Baseline detector (non‑deep learning)\n",
"\n",
"A lightweight baseline is included for context. \n",
"Each window is scored with a robust z‑score built from the window median and MAD (median absolute deviation), then aggregated with a max over timesteps/features. No model training is involved.\n",
"\n",
"The baseline is expected to underperform on multivariate drift because it does not learn cross‑feature structure.\n"
]
},
{
"cell_type": "code",
"execution_count": 27,
"id": "3b9b74a5",
"metadata": {},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:45:33.317189: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"Baseline threshold (p99) on clean validation: 89.46832092285166\n",
"=== Baseline - spike (MAD max-z) ===\n",
"AP (PR-AUC): 0.7663836164906992\n",
"Precision: 0.9820962327489743\n",
"Recall: 0.16714276645718276\n",
"F1: 0.28566778778344365\n",
"Confusion matrix [[TN, FP],[FN, TP]]:\n",
"[[47139 48]\n",
" [13120 2633]]\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:45:37.942597: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n"
]
},
{
"data": {
"image/png": "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",
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"=== Baseline - drift (MAD max-z) ===\n",
"AP (PR-AUC): 0.24885421382448747\n",
"Precision: 0.2916666666666667\n",
"Recall: 0.0013315579227696406\n",
"F1: 0.0026510130657072524\n",
"Confusion matrix [[TN, FP],[FN, TP]]:\n",
"[[47118 51]\n",
" [15750 21]]\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:45:42.676896: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n"
]
},
{
"data": {
"image/png": "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",
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"# Window-based robust baseline using Median Absolute Deviation (MAD).\n",
"# The baseline produces a scalar anomaly score per window without model training.\n",
"\n",
"def baseline_mad_window_scores(ds_windows: tf.data.Dataset) -> np.ndarray:\n",
" # Score definition:\n",
" # - per window (L,F), compute per-feature median over time\n",
" # - compute per-feature MAD over time\n",
" # - compute robust z = |x - median| / (MAD + eps)\n",
" # - window score = max robust z over all timesteps and features\n",
" scores = []\n",
" for batch in ds_windows:\n",
" x = batch.numpy() # (B,L,F)\n",
" med = np.median(x, axis=1, keepdims=True) # (B,1,F)\n",
" mad = np.median(np.abs(x - med), axis=1, keepdims=True) + 1e-6\n",
" z = np.abs(x - med) / mad\n",
" s = np.max(z, axis=(1, 2)) # (B,)\n",
" scores.append(s)\n",
" return np.concatenate(scores, axis=0)\n",
"\n",
"# Threshold from clean validation windows (p99)\n",
"baseline_val_scores = baseline_mad_window_scores(ds_val)\n",
"baseline_thr = float(np.percentile(baseline_val_scores, 99.0))\n",
"\n",
"# Evaluate on deterministic injected datasets\n",
"def baseline_eval(ds_injected: tf.data.Dataset, threshold: float):\n",
" y_true = []\n",
" y_score = []\n",
" y_pred = []\n",
" for x_batch, y_batch, _t0 in ds_injected:\n",
" x = x_batch.numpy()\n",
" med = np.median(x, axis=1, keepdims=True)\n",
" mad = np.median(np.abs(x - med), axis=1, keepdims=True) + 1e-6\n",
" z = np.abs(x - med) / mad\n",
" score = np.max(z, axis=(1, 2))\n",
" y = y_batch.numpy().astype(int)\n",
"\n",
" y_true.append(y)\n",
" y_score.append(score)\n",
" y_pred.append((score > threshold).astype(int))\n",
"\n",
" y_true = np.concatenate(y_true)\n",
" y_score = np.concatenate(y_score)\n",
" y_pred = np.concatenate(y_pred)\n",
"\n",
" ap = average_precision_score(y_true, y_score) if len(np.unique(y_true)) > 1 else np.nan\n",
" p, r, f1, _ = precision_recall_fscore_support(y_true, y_pred, average=\"binary\", zero_division=0)\n",
" cm = confusion_matrix(y_true, y_pred, labels=[0, 1])\n",
" return {\"ap\": float(ap), \"precision\": float(p), \"recall\": float(r), \"f1\": float(f1), \"cm\": cm}\n",
"\n",
"print(\"Baseline threshold (p99) on clean validation:\", baseline_thr)\n",
"\n",
"m_base_spike = baseline_eval(make_injected_dataset(ds_test, \"spike\"), baseline_thr)\n",
"print(\"=== Baseline - spike (MAD max-z) ===\")\n",
"print(\"AP (PR-AUC):\", m_base_spike[\"ap\"])\n",
"print(\"Precision:\", m_base_spike[\"precision\"])\n",
"print(\"Recall:\", m_base_spike[\"recall\"])\n",
"print(\"F1:\", m_base_spike[\"f1\"])\n",
"print(\"Confusion matrix [[TN, FP],[FN, TP]]:\")\n",
"print(m_base_spike[\"cm\"])\n",
"plot_confusion_matrix(m_base_spike[\"cm\"], \"Baseline — Spike (MAD)\")\n",
"\n",
"m_base_drift = baseline_eval(make_injected_dataset(ds_test, \"drift\"), baseline_thr)\n",
"print(\"=== Baseline - drift (MAD max-z) ===\")\n",
"print(\"AP (PR-AUC):\", m_base_drift[\"ap\"])\n",
"print(\"Precision:\", m_base_drift[\"precision\"])\n",
"print(\"Recall:\", m_base_drift[\"recall\"])\n",
"print(\"F1:\", m_base_drift[\"f1\"])\n",
"print(\"Confusion matrix [[TN, FP],[FN, TP]]:\")\n",
"print(m_base_drift[\"cm\"])\n",
"plot_confusion_matrix(m_base_drift[\"cm\"], \"Baseline — Drift (MAD)\")"
]
},
{
"cell_type": "markdown",
"id": "a59b5ed4",
"metadata": {},
"source": [
"## PCA Reconstruction-Error Baseline\n",
"\n",
"To strengthen the baseline comparison, a PCA reconstruction-error baseline is added in addition to the MAD baseline. This provides a **learned linear reconstruction model**, which is more comparable to the autoencoder framework than a purely statistical thresholding method.\n",
"\n",
"Each window has shape `(144, 14)`, so for PCA each window is flattened into a vector of length `2016`. PCA is then fitted on **clean training windows only**, and reconstruction error is used as the anomaly score.\n",
"\n",
"To keep the comparison controlled, the PCA model uses **32 components**, matching the latent size used in the CNN and LSTM autoencoders.\n"
]
},
{
"cell_type": "code",
"execution_count": 28,
"id": "76493d27",
"metadata": {},
"outputs": [],
"source": [
"# Clean, non-shuffled datasets for PCA fitting/evaluation\n",
"ds_train_pca = make_window_dataset(X_train, WINDOW, SHIFT_TRAIN, BATCH_SIZE, shuffle=False)\n",
"ds_val_pca = make_window_dataset(X_val, WINDOW, SHIFT_EVAL, BATCH_SIZE, shuffle=False)\n",
"ds_test_pca = make_window_dataset(X_test, WINDOW, SHIFT_EVAL, BATCH_SIZE, shuffle=False)\n",
"\n",
"# Use 32 components to mirror the AE latent size\n",
"PCA_COMPONENTS = 32\n",
"ipca = IncrementalPCA(n_components=PCA_COMPONENTS)"
]
},
{
"cell_type": "markdown",
"id": "4c13294b",
"metadata": {},
"source": [
"## Fit PCA on Clean Training Windows\n",
"\n",
"Incremental PCA is fitted on the flattened clean training windows. Incremental PCA is used instead of standard PCA so that fitting can proceed batch by batch, which is more memory-efficient for a large number of windows.\n"
]
},
{
"cell_type": "code",
"execution_count": 29,
"id": "1916f44a",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"PCA fitted with n_components = 32\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:45:49.062448: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n"
]
}
],
"source": [
"# Fit Incremental PCA on flattened clean training windows\n",
"for batch in ds_train_pca:\n",
" x = batch.numpy().reshape(batch.shape[0], -1) # (B, 144*14)\n",
" ipca.partial_fit(x)\n",
"\n",
"print(\"PCA fitted with n_components =\", PCA_COMPONENTS)"
]
},
{
"cell_type": "markdown",
"id": "915ec34f",
"metadata": {},
"source": [
"## PCA Reconstruction Scores\n",
"\n",
"After fitting PCA, each window is projected into the learned low-dimensional subspace and reconstructed back to the original flattened form. The absolute reconstruction error is then computed.\n",
"\n",
"Two score types are used, matching the logic of the autoencoder evaluation:\n",
"\n",
"- **Mean reconstruction error** for drift, because drift is distributed across many timesteps\n",
"- **Max reconstruction error** for spikes, because spikes are localized\n"
]
},
{
"cell_type": "code",
"execution_count": 30,
"id": "51149b09",
"metadata": {},
"outputs": [],
"source": [
"def _pca_reconstruct_batch(ipca, x_batch: tf.Tensor):\n",
" x = x_batch.numpy().reshape(x_batch.shape[0], -1) # (B, L*F)\n",
" x_rec = ipca.inverse_transform(ipca.transform(x)) # (B, L*F)\n",
" return x, x_rec\n",
"\n",
"def pca_window_scores_mean(ipca, ds_windows: tf.data.Dataset) -> np.ndarray:\n",
" scores = []\n",
" for batch in ds_windows:\n",
" x, x_rec = _pca_reconstruct_batch(ipca, batch)\n",
" err = np.abs(x - x_rec)\n",
" scores.append(err.mean(axis=1))\n",
" return np.concatenate(scores, axis=0)\n",
"\n",
"def pca_window_scores_max(ipca, ds_windows: tf.data.Dataset) -> np.ndarray:\n",
" scores = []\n",
" for batch in ds_windows:\n",
" x, x_rec = _pca_reconstruct_batch(ipca, batch)\n",
" err = np.abs(x - x_rec)\n",
" scores.append(err.max(axis=1))\n",
" return np.concatenate(scores, axis=0)\n",
"\n",
"def pca_eval_on_injected(ipca, ds_injected: tf.data.Dataset, threshold: float, score_mode: str = \"mean\"):\n",
" y_true = []\n",
" y_score = []\n",
" y_pred = []\n",
"\n",
" for x_batch, y_batch, _t0 in ds_injected:\n",
" x, x_rec = _pca_reconstruct_batch(ipca, x_batch)\n",
" err = np.abs(x - x_rec)\n",
"\n",
" if score_mode == \"mean\":\n",
" score = err.mean(axis=1)\n",
" elif score_mode == \"max\":\n",
" score = err.max(axis=1)\n",
" else:\n",
" raise ValueError(\"score_mode must be 'mean' or 'max'\")\n",
"\n",
" y = y_batch.numpy().astype(int)\n",
"\n",
" y_true.append(y)\n",
" y_score.append(score)\n",
" y_pred.append((score > threshold).astype(int))\n",
"\n",
" y_true = np.concatenate(y_true)\n",
" y_score = np.concatenate(y_score)\n",
" y_pred = np.concatenate(y_pred)\n",
"\n",
" ap = average_precision_score(y_true, y_score) if len(np.unique(y_true)) > 1 else np.nan\n",
" p, r, f1, _ = precision_recall_fscore_support(y_true, y_pred, average=\"binary\", zero_division=0)\n",
" cm = confusion_matrix(y_true, y_pred, labels=[0, 1])\n",
"\n",
" return {\"ap\": float(ap), \"precision\": float(p), \"recall\": float(r), \"f1\": float(f1), \"cm\": cm}"
]
},
{
"cell_type": "markdown",
"id": "0fddb97e",
"metadata": {},
"source": [
"## Threshold Calibration for PCA\n",
"\n",
"As with the deep models, anomaly thresholds are calibrated on **clean validation windows** using the 99th percentile of reconstruction scores.\n",
"\n",
"This keeps the evaluation pipeline consistent:\n",
"\n",
"- **p99 of mean score** for drift detection\n",
"- **p99 of max score** for spike detection\n"
]
},
{
"cell_type": "code",
"execution_count": 31,
"id": "ebff57bb",
"metadata": {},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:46:18.784200: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"PCA threshold mean (p99): 0.24854606704548501\n",
"PCA threshold max (p99): 3.935686568496929\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:47:02.766750: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n"
]
}
],
"source": [
"pca_thr_mean = percentile_threshold(pca_window_scores_mean(ipca, ds_val_pca), 99.0)\n",
"pca_thr_max = percentile_threshold(pca_window_scores_max(ipca, ds_val_pca), 99.0)\n",
"\n",
"print(\"PCA threshold mean (p99):\", pca_thr_mean)\n",
"print(\"PCA threshold max (p99):\", pca_thr_max)"
]
},
{
"cell_type": "markdown",
"id": "5008ce6e",
"metadata": {},
"source": [
"## Evaluate PCA on Injected Spike and Drift Anomalies\n",
"\n",
"The PCA baseline is evaluated on the same deterministically injected test datasets used for the autoencoders. This ensures that the comparison remains fair across all methods.\n",
"\n",
"The resulting metrics include:\n",
"\n",
"- Average Precision (AP / PR-AUC)\n",
"- Precision\n",
"- Recall\n",
"- F1-score\n",
"- Confusion matrix\n"
]
},
{
"cell_type": "code",
"execution_count": 32,
"id": "33a20699",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"=== PCA - spike (max score) ===\n",
"AP (PR-AUC): 0.9999547036053935\n",
"Precision: 1.0\n",
"Recall: 0.9994286802513807\n",
"F1: 0.9997142585008096\n",
"Confusion matrix [[TN, FP],[FN, TP]]:\n",
"[[47187 0]\n",
" [ 9 15744]]\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:47:16.811206: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n"
]
},
{
"data": {
"image/png": "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",
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n",
"=== PCA - drift (mean score) ===\n",
"AP (PR-AUC): 0.9207203783563729\n",
"Precision: 0.9921122839577775\n",
"Recall: 0.542324519688035\n",
"F1: 0.7012955067235159\n",
"Confusion matrix [[TN, FP],[FN, TP]]:\n",
"[[47101 68]\n",
" [ 7218 8553]]\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-03-17 15:47:30.704862: W tensorflow/core/framework/local_rendezvous.cc:404] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n"
]
},
{
"data": {
"image/png": "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",
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"print(\"=== PCA - spike (max score) ===\")\n",
"m_pca_spike = pca_eval_on_injected(\n",
" ipca,\n",
" make_injected_dataset(ds_test_pca, \"spike\"),\n",
" pca_thr_max,\n",
" score_mode=\"max\"\n",
")\n",
"print(\"AP (PR-AUC):\", m_pca_spike[\"ap\"])\n",
"print(\"Precision:\", m_pca_spike[\"precision\"])\n",
"print(\"Recall:\", m_pca_spike[\"recall\"])\n",
"print(\"F1:\", m_pca_spike[\"f1\"])\n",
"print(\"Confusion matrix [[TN, FP],[FN, TP]]:\")\n",
"print(m_pca_spike[\"cm\"])\n",
"plot_confusion_matrix(m_pca_spike[\"cm\"], \"PCA — Spike (max score)\", cmap=\"Blues\")\n",
"\n",
"print(\"\\n=== PCA - drift (mean score) ===\")\n",
"m_pca_drift = pca_eval_on_injected(\n",
" ipca,\n",
" make_injected_dataset(ds_test_pca, \"drift\"),\n",
" pca_thr_mean,\n",
" score_mode=\"mean\"\n",
")\n",
"print(\"AP (PR-AUC):\", m_pca_drift[\"ap\"])\n",
"print(\"Precision:\", m_pca_drift[\"precision\"])\n",
"print(\"Recall:\", m_pca_drift[\"recall\"])\n",
"print(\"F1:\", m_pca_drift[\"f1\"])\n",
"print(\"Confusion matrix [[TN, FP],[FN, TP]]:\")\n",
"print(m_pca_drift[\"cm\"])\n",
"plot_confusion_matrix(m_pca_drift[\"cm\"], \"PCA — Drift (mean score)\", cmap=\"Blues\")\n"
]
},
{
"cell_type": "markdown",
"id": "9c538564",
"metadata": {},
"source": [
"## Interpretation of PCA Baseline Results\n",
"\n",
"The PCA baseline provides a stronger classical comparison than MAD.\n",
"\n",
"The results show a clear pattern:\n",
"\n",
"- **Spike detection is extremely strong**, indicating that highly localized anomalies can already be captured very well by linear reconstruction.\n",
"- **Drift detection is clearly weaker than the CNN and LSTM**, suggesting that gradual anomalies benefit more from nonlinear temporal models.\n",
"\n",
"This means the strongest added value of the deep temporal models appears in the detection of **gradual drift**, rather than in sharp localized spikes.\n"
]
},
{
"cell_type": "markdown",
"id": "e5d33825",
"metadata": {},
"source": [
"## Summary, Discussion, Limitations, and Conclusion\n",
"\n",
"This investigation compared a **1D-CNN autoencoder** and an **LSTM autoencoder** for anomaly detection in multivariate windows from the **Jena Climate** dataset, using a denoising autoencoder setup and synthetic spike/drift injections for controlled evaluation.\n",
"\n",
"The results show that both autoencoders were effective, but the **CNN autoencoder was the strongest deep model overall** under the chosen benchmark. For **drift detection**, the CNN achieved the best overall classification performance, while the **LSTM reacted earlier** to gradual change, showing a trade-off between **overall accuracy** and **detection timeliness**.\n",
"\n",
"The stronger baseline comparison also clarified the picture. The **MAD baseline** was too weak for this task, especially for drift. The **PCA reconstruction-error baseline** was added as a result. It proved to be much stronger and was near-perfect for **spike detection**, suggesting that sharp localized anomalies can be captured very well even by a linear reconstruction model. However, PCA was clearly weaker than the deep temporal models for **drift**, which indicates that the added value of deep learning in this project appears most strongly for **gradual, distributed anomalies**.\n",
"\n",
"These findings should still be interpreted with some caution. The anomalies were **synthetic**, hyperparameter tuning was **targeted rather than exhaustive**, and **anomaly magnitude** itself can strongly influence reconstruction error and therefore the final evaluation metrics. Even so, the relative comparison between methods remains informative because all models were tested under the same controlled setup.\n",
"\n",
"Overall, the main conclusion is that **deep temporal models are most useful when anomaly structure unfolds over time**, while simpler reconstruction methods may already be sufficient for strong localized spikes. Future work could include a sensitivity study over anomaly magnitudes, additional anomaly types, a dense MLP autoencoder baseline, and evaluation on real labeled anomalies."
]
},
{
"cell_type": "markdown",
"id": "f61e568b",
"metadata": {},
"source": [
"## 14. Saving results\n",
"\n",
"Exporting results as CSV supports reproducibility and clean presentation.\n"
]
},
{
"cell_type": "code",
"execution_count": 33,
"id": "95baafed",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Saved: results/run_metadata.json\n",
"Artifacts: {'window_justify_micro': 'results/window_justify_micro.csv', 'stride_justify_micro': 'results/stride_justify_micro.csv'}\n"
]
}
],
"source": [
"import os\n",
"import json\n",
"import platform\n",
"import hashlib\n",
"from pathlib import Path\n",
"\n",
"out_dir = Path(\"results\")\n",
"out_dir.mkdir(exist_ok=True)\n",
"\n",
"def sha256_file(path: str) -> str:\n",
" h = hashlib.sha256()\n",
" with open(path, \"rb\") as f:\n",
" for chunk in iter(lambda: f.read(1024 * 1024), b\"\"):\n",
" h.update(chunk)\n",
" return h.hexdigest()\n",
"\n",
"artifacts = {}\n",
"\n",
"# Save compact justification tables if present\n",
"if \"win_df\" in globals():\n",
" p = out_dir / \"window_justify_micro.csv\"\n",
" win_df.to_csv(p, index=False)\n",
" artifacts[\"window_justify_micro\"] = str(p)\n",
"\n",
"if \"st_df\" in globals():\n",
" p = out_dir / \"stride_justify_micro.csv\"\n",
" st_df.to_csv(p, index=False)\n",
" artifacts[\"stride_justify_micro\"] = str(p)\n",
"\n",
"# Save main evaluation metrics if present\n",
"main_metrics = {}\n",
"for name in [\"m_cnn_spike\", \"m_lstm_spike\", \"m_cnn_drift\", \"m_lstm_drift\", \"m_base_spike\", \"m_base_drift\"]:\n",
" if name in globals():\n",
" m = globals()[name]\n",
" main_metrics[name] = {\n",
" \"ap\": m.get(\"ap\", None),\n",
" \"precision\": m.get(\"precision\", None),\n",
" \"recall\": m.get(\"recall\", None),\n",
" \"f1\": m.get(\"f1\", None),\n",
" \"cm\": m.get(\"cm\", None).tolist() if isinstance(m.get(\"cm\", None), np.ndarray) else m.get(\"cm\", None),\n",
" }\n",
"\n",
"csv_path_used = globals().get(\"CSV_PATH\", globals().get(\"csv_path\", None))\n",
"csv_hash = sha256_file(csv_path_used) if (csv_path_used and os.path.exists(csv_path_used)) else None\n",
"\n",
"config = {\n",
" \"train_view\": TRAIN_VIEW,\n",
" \"noise_std\": float(NOISE_STD),\n",
" \"window\": int(WINDOW),\n",
" \"shift_train\": int(SHIFT_TRAIN),\n",
" \"shift_eval\": int(SHIFT_EVAL),\n",
" \"batch_size\": int(BATCH_SIZE),\n",
" \"tensorflow_version\": tf.__version__,\n",
" \"keras_version\": keras.__version__,\n",
" \"python_version\": platform.python_version(),\n",
" \"platform\": platform.platform(),\n",
" \"seed\": int(SEED),\n",
" \"csv_path\": csv_path_used,\n",
" \"csv_sha256\": csv_hash,\n",
" \"deterministic_ops\": True,\n",
" \"devices\": {\n",
" \"gpus\": [d.name for d in tf.config.list_physical_devices(\"GPU\")],\n",
" \"cpus\": [d.name for d in tf.config.list_physical_devices(\"CPU\")],\n",
" },\n",
"}\n",
"\n",
"meta = {\"config\": config, \"main_metrics\": main_metrics, \"artifacts\": artifacts}\n",
"\n",
"meta_path = out_dir / \"run_metadata.json\"\n",
"meta_path.write_text(json.dumps(meta, indent=2))\n",
"print(\"Saved:\", meta_path)\n",
"print(\"Artifacts:\", artifacts)\n"
]
},
{
"cell_type": "markdown",
"id": "3f55aabe",
"metadata": {},
"source": [
"## 15. References\n",
"\n",
"- Chollet, F. (2021). Deep Learning with Python (2nd ed.). Manning Publications.\n",
"\n",
"- Géron, A. (2022). Hands-On Machine Learning with Scikit-Learn, Keras, and TensorFlow: Concepts, Tools, and Techniques to Build Intelligent Systems (3rd ed.). O'Reilly Media.\n",
"\n",
"- Keras. (2020, May 31). Timeseries anomaly detection using an autoencoder. https://keras.io/examples/timeseries/timeseries_anomaly_detection/\n",
"\n",
"- Keras. (2020, June 23). Timeseries forecasting for weather prediction. https://keras.io/examples/timeseries/timeseries_weather_forecasting/\n",
"\n",
"- Max Planck Institute for Biogeochemistry. (n.d.). Jena Climate Dataset [Data set]. https://storage.googleapis.com/tensorflow/tf-keras-datasets/jena_climate_2009_2016.csv.zip\n",
"\n",
"- scikit-learn developers. (n.d.). PCA. scikit-learn. https://scikit-learn.org/stable/modules/generated/sklearn.decomposition.PCA.html"
]
}
],
"metadata": {
"kernelspec": {
"display_name": "Python 3",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.12.6"
}
},
"nbformat": 4,
"nbformat_minor": 5
}