Adding reconstruction error section

DiogenesAnalytics · Jan 2, 2024 · 9e7ae74 · 9e7ae74
1 parent c0b983a
commit 9e7ae74
Show file tree

Hide file tree

Showing 2 changed files with 91 additions and 15 deletions.
diff --git a/notebooks/demo/mnist_dataset.ipynb b/notebooks/demo/mnist_dataset.ipynb
@@ -48,6 +48,7 @@
    "outputs": [],
    "source": [
     "# get necessary libs for data/preprocessing\n",
+    "import tensorflow as tf\n",
     "from keras.datasets import mnist\n",
     "\n",
     "# load the data\n",
@@ -56,8 +57,18 @@
     "# preprocess the data (normalize)\n",
     "x_train = x_train.astype(\"float32\") / 255.\n",
     "x_test = x_test.astype(\"float32\") / 255.\n",
-    "print(f\"Train shape: {x_train.shape}\")\n",
-    "print(f\"Test shape: {x_test.shape}\")"
+    "\n",
+    "# convert to tf datasets\n",
+    "train_ds = tf.data.Dataset.from_tensor_slices((x_train, x_train))\n",
+    "test_ds = tf.data.Dataset.from_tensor_slices((x_test, x_test))\n",
+    "\n",
+    "# set a few params\n",
+    "BATCH_SIZE = 64\n",
+    "SHUFFLE_BUFFER_SIZE = 100\n",
+    "\n",
+    "# update with batch/buffer size\n",
+    "train_ds = train_ds.shuffle(SHUFFLE_BUFFER_SIZE).batch(BATCH_SIZE)\n",
+    "test_ds = test_ds.batch(BATCH_SIZE)"
    ]
   },
   {
@@ -77,12 +88,20 @@
    "outputs": [],
    "source": [
     "# get libs for training ae\n",
-    "from autoencoder.model.minimal import Min2DAE\n",
+    "from autoencoder.model.minimal import Min2DAE, Min2DParams\n",
+    "\n",
+    "# seupt config\n",
+    "config = Min2DParams(\n",
+    "    l0={\"input_shape\": (28, 28)},\n",
+    "    l2={\"units\": 32 * 1},\n",
+    "    l3={\"units\": 28 * 28 * 1},\n",
+    "    l4={\"target_shape\": (28, 28)},\n",
+    ")\n",
     "\n",
     "# get ae instance\n",
-    "autoencoder = Min2DAE()\n",
+    "autoencoder = Min2DAE(config)\n",
     "\n",
-    "# check model topology\n",
+    "# check network topology\n",
     "autoencoder.summary()"
    ]
   },
@@ -103,13 +122,10 @@
     "autoencoder.compile(optimizer=\"adam\", loss=\"binary_crossentropy\")\n",
     "\n",
     "# begin model fit\n",
-    "history = autoencoder.fit(\n",
-    "    x=x_train,\n",
-    "    y=x_train,\n",
+    "autoencoder.fit(\n",
+    "    x=train_ds,\n",
     "    epochs=50,\n",
-    "    batch_size=256,\n",
-    "    shuffle=True,\n",
-    "    validation_data=(x_test, x_test),\n",
+    "    validation_data=test_ds,\n",
     "    callbacks=[early_stop_callback],\n",
     ")"
    ]
@@ -137,18 +153,50 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "1ea208f6-b50a-405d-b07a-ad86736c0a43",
+   "id": "d14a4180-ca54-437d-abc4-21427f45eed2",
    "metadata": {},
    "outputs": [],
    "source": [
     "# get viz func\n",
     "from autoencoder.data import compare_image_predictions\n",
     "\n",
-    "# get decoded images\n",
-    "decoded_imgs = autoencoder.predict(x=x_train)\n",
+    "# get samples from validation dataset\n",
+    "val_samples = test_ds.take(1)\n",
+    "\n",
+    "# get raw numpy arrays\n",
+    "val_input = [item for pair in val_samples.as_numpy_iterator() for item in pair[0]]\n",
+    "\n",
+    "# and decoded\n",
+    "decoded_imgs = autoencoder.predict(x=val_samples)\n",
     "\n",
     "# display\n",
-    "compare_image_predictions(x_train, decoded_imgs)"
+    "compare_image_predictions(val_input, decoded_imgs)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "e299de94-d802-4d3d-82c5-875ddde80232",
+   "metadata": {},
+   "source": [
+    "## Reconstruction Error Distribution\n",
+    "Now let us take peak into this dataset and see how well the *autoencoder* is capturing the image features."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "60a9bac7-2902-4ea7-9c06-216584369c6d",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# get custom class\n",
+    "from autoencoder.data.evaluate import AutoencoderEvaluator\n",
+    "\n",
+    "# get instance\n",
+    "ae_eval = AutoencoderEvaluator(autoencoder, test_ds, axis=(1, 2))\n",
+    "\n",
+    "# view distribution\n",
+    "ae_eval.view_error_distribution(\"MNIST Autoencoder: Reconstruction Error Distribution\")"
    ]
   }
  ],

diff --git a/notebooks/demo/tf_flowers.ipynb b/notebooks/demo/tf_flowers.ipynb
@@ -191,6 +191,34 @@
     "# display\n",
     "compare_image_predictions(val_input, decoded_imgs)"
    ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "672e0631-950c-44f8-9f7a-9268062c1a55",
+   "metadata": {},
+   "source": [
+    "## Reconstruction Error Distribution\n",
+    "Now let us take peak into this dataset and see how well the *autoencoder* is capturing the image features."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "60a9bac7-2902-4ea7-9c06-216584369c6d",
+   "metadata": {
+    "scrolled": true
+   },
+   "outputs": [],
+   "source": [
+    "# get custom class\n",
+    "from autoencoder.data.evaluate import AutoencoderEvaluator\n",
+    "\n",
+    "# get instance\n",
+    "ae_eval = AutoencoderEvaluator(autoencoder, x_val)\n",
+    "\n",
+    "# view distribution\n",
+    "ae_eval.view_error_distribution(\"tf_flowers Autoencoder: Reconstruction Error Distribution\")"
+   ]
   }
  ],
  "metadata": {