2025 AI-assisted Update

notebooks/dl/dl-cnn-image-classification.ipynb

@@ -140,7 +140,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "seaborn.distplot(img_a.flatten())"
+    "seaborn.histplot(img_a.flatten(), kde=True)\n"
    ]
   },
   {
@@ -205,7 +205,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "seaborn.distplot(img_a_conv.flatten())"
+    "seaborn.histplot(img_a_conv.flatten(), kde=True)\n"
    ]
   },
   {
@@ -471,15 +471,15 @@
     "\n",
     "net = keras.models.Sequential(\n",
     "    [\n",
-    "        keras.layers.Convolution2D(\n",
+    "        keras.layers.Conv2D(\n",
     "            filters=n1,\n",
     "            kernel_size=(3,3),\n",
     "            activation=\"relu\",\n",
     "        ),\n",
     "        keras.layers.MaxPooling2D(\n",
     "            pool_size=(2,2)\n",
     "        ),\n",
-    "        keras.layers.Convolution2D(\n",
+    "        keras.layers.Conv2D(\n",
     "            filters=n2,\n",
     "            kernel_size=(3,3),\n",
     "            activation=\"relu\"\n",

notebooks/dl/dl-explainable.ipynb

@@ -115,13 +115,16 @@
    "metadata": {},
    "outputs": [],
    "source": [
+    "from tensorflow.keras.utils import load_img, img_to_array\n",
+    "\n",
+    "\n",
     "def preprocess_image(img_path):\n",
     "    \"\"\"Preprocess the \"\"\"\n",
-    "    img = keras.preprocessing.image.load_img(\n",
+    "    img = load_img(\n",
     "        img_path, \n",
     "        target_size=(299, 299)\n",
     "    )\n",
-    "    img_array = keras.preprocessing.image.img_to_array(img)\n",
+    "    img_array = img_to_array(img)\n",
     "    #img_array = numpy.expand_dims(img_array, axis=0)\n",
     "    img_array = inception_v3.preprocess_input(img_array)\n",
     "    img_array = img_array / 2 + 0.5  # brighten\n",

notebooks/dl/dl-house-price-regression.ipynb

@@ -493,12 +493,14 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "This function is now passed to a wrapper class that aims to implement the `sklearn` estimator interface. We can now pass the data in the shape usual for `sklearn` to train the classifier..."
+    "This function is now passed to a wrapper class that aims to implement the `sklearn` estimator interface.\n",
+    "\n",
+    "First, we need to import the `KerasRegressor` from `scikeras`. We can now pass the data in the shape usual for `sklearn` to train the regressor..."
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": 17,
+   "execution_count": null,
    "metadata": {},
    "outputs": [
     {
@@ -532,15 +534,17 @@
     }
    ],
    "source": [
-    "model = keras.wrappers.scikit_learn.KerasRegressor(\n",
-    "    build_fn=build_regressor_network,\n",
+    "from scikeras.wrappers import KerasRegressor\n",
+    "\n",
+    "model = KerasRegressor(\n",
+    "    model=build_regressor_network,\n",
     "    epochs=25, \n",
     "    batch_size=10,\n",
     ")\n",
     "\n",
     "history = model.fit(\n",
-    "    x=X_train,\n",
-    "    y=y_train,\n",
+    "    X_train,\n",
+    "    y_train,\n",
     "    validation_split=0.2\n",
     ")"
    ]
@@ -559,7 +563,7 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "... and ask for class predictions via the `predict` method. However, we get the predictions in the nested array shape that is usual for TensorFlow, so we need to flatten the array to get a 1D-vector."
+    "... and ask for predictions via the `predict` method:"
    ]
   },
   {
@@ -568,7 +572,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "y_pred = model.predict(X_test).flatten()"
+    "y_pred = model.predict(X_test)"
    ]
   },
   {
@@ -577,8 +581,8 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "seaborn.distplot(data[target_col])\n",
-    "seaborn.distplot(y_pred)\n",
+    "seaborn.histplot(data[target_col], kde=True)\n",
+    "seaborn.histplot(y_pred, kde=True)\n",
     "plt.title(\"distribution of target variable and predictions\")"
    ]
   },
@@ -588,7 +592,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "seaborn.distplot(y_pred - y_test)\n",
+    "seaborn.histplot(y_pred - y_test, kde=True)\n",
     "plt.title(\"distribution of error\")"
    ]
   },

notebooks/dl/dl-iris-classifier.ipynb

@@ -212,7 +212,9 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "This function is now passed to a wrapper class that aims to implement the `sklearn` estimator interface:"
+    "This function is now passed to a wrapper class that aims to implement the `sklearn` estimator interface.\n",
+    "\n",
+    "First, we need to import the `KerasClassifier` from `scikeras`:"
    ]
   },
   {
@@ -221,8 +223,10 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "keras.wrappers.scikit_learn.KerasClassifier(\n",
-    "    build_fn=build_binary_classifier,\n",
+    "from scikeras.wrappers import KerasClassifier\n",
+    "\n",
+    "KerasClassifier(\n",
+    "    model=build_binary_classifier,\n",
     "    epochs=100, \n",
     "    batch_size=5,\n",
     ")\n"
@@ -241,22 +245,22 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "model = keras.wrappers.scikit_learn.KerasClassifier(\n",
-    "    build_fn=build_binary_classifier,\n",
+    "model = KerasClassifier(\n",
+    "    model=build_binary_classifier,\n",
     "    epochs=20, \n",
     "    batch_size=5,\n",
     ")\n",
     "model.fit(\n",
-    "    x=X_train,\n",
-    "    y=y_train\n",
+    "    X_train,\n",
+    "    y_train\n",
     ")"
    ]
   },
   {
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "... and ask for class predictions via the `predict` method. However, we get the predictions in the nested array shape that is usual for TensorFlow, so we need to flatten the array to get a 1D-vector."
+    "... and ask for class predictions via the `predict` method:"
    ]
   },
   {
@@ -265,7 +269,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "y_pred = model.predict(X_test).flatten()"
+    "y_pred = model.predict(X_test)"
    ]
   },
   {

notebooks/dl/dl-learning-process.ipynb

@@ -237,7 +237,7 @@
     "\n",
     "net = keras.models.Sequential(\n",
     "    [\n",
-    "        keras.layers.Convolution2D(\n",
+    "        keras.layers.Conv2D(\n",
     "            input_shape=input_shape,\n",
     "            filters=n1,\n",
     "            kernel_size=(5,5),\n",
@@ -246,7 +246,7 @@
     "        keras.layers.MaxPooling2D(\n",
     "            pool_size=(2,2)\n",
     "        ),\n",
-    "        keras.layers.Convolution2D(\n",
+    "        keras.layers.Conv2D(\n",
     "            filters=n2,\n",
     "            kernel_size=(5,5),\n",
     "            activation=\"relu\"\n",

notebooks/dl/dl-pretrained.ipynb

@@ -175,7 +175,8 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "img_a = keras.preprocessing.image.img_to_array(img)\n",
+    "from tensorflow.keras.utils import img_to_array\n",
+    "img_a = img_to_array(img)\n",
     "img_a.shape"
    ]
   },
@@ -185,7 +186,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "seaborn.distplot(img_a.flatten())"
+    "seaborn.histplot(img_a.flatten(), kde=True)\n"
    ]
   },
   {
@@ -211,7 +212,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "seaborn.distplot(img_a.flatten())"
+    "seaborn.histplot(img_a.flatten(), kde=True)\n"
    ]
   },
   {
@@ -318,7 +319,7 @@
     "        response = requests.get(img_url)\n",
     "        img = PIL.Image.open(io.BytesIO(response.content))\n",
     "        img = img.resize(inception_dim)\n",
-    "        img_a = keras.preprocessing.image.img_to_array(img)\n",
+    "        img_a = img_to_array(img)\n",
     "        img_a = inception_v3.preprocess_input(img_a)\n",
     "        predictions = self.model.predict(\n",
     "            numpy.array([img_a])\n",

notebooks/dl/dl-save-models.ipynb

@@ -118,7 +118,7 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "This method of the Keras model saves the network - including its weights - to a [HDF5](https://en.m.wikipedia.org/wiki/Hierarchical_Data_Format) file:"
+    "This method of the Keras model saves the network - including its weights - to a file using the native Keras v3 format (`.keras`):"
    ]
   },
   {
@@ -127,7 +127,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "net.save(\"saved-model.h5\")"
+    "net.save(\"saved-model.keras\")"
    ]
   },
   {
@@ -136,7 +136,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "!ls -lah saved-model.h5"
+    "!ls -lah saved-model.keras"
    ]
   },
   {
@@ -152,7 +152,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "net_reloaded = keras.models.load_model(\"saved-model.h5\")"
+    "net_reloaded = keras.models.load_model(\"saved-model.keras\")"
    ]
   },
   {
@@ -232,7 +232,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "!rm saved-model.h5"
+    "!rm saved-model.keras"
    ]
   },
   {

notebooks/exercises/ml-exercise-house-prices.ipynb

@@ -929,9 +929,9 @@
    "source": [
     "if ready:\n",
     "    predicted_pd = predicted[[\"SalePrice\", \"prediction\"]]\n",
-    "    seaborn.distplot(predicted_pd[\"SalePrice\"])\n",
-    "    seaborn.distplot(predicted_pd[\"prediction\"])\n",
-    "    seaborn.distplot(predicted_pd[\"SalePrice\"] - predicted_pd[\"prediction\"])    "
+    "    seaborn.histplot(predicted_pd[\"SalePrice\"], kde=True)\n",
+    "    seaborn.histplot(predicted_pd[\"prediction\"], kde=True)\n",
+    "    seaborn.histplot(predicted_pd[\"SalePrice\"] - predicted_pd[\"prediction\"], kde=True)\n"
    ]
   },
   {
@@ -964,4 +964,4 @@
  },
  "nbformat": 4,
  "nbformat_minor": 4
-}
+}

notebooks/exercises/spark-exercise-house-prices-solution.ipynb

@@ -316,7 +316,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "seaborn.distplot(predicted_pd[\"SalePrice\"])"
+    "seaborn.histplot(predicted_pd[\"SalePrice\"], kde=True)\n"
    ]
   },
   {
@@ -325,7 +325,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "seaborn.distplot(predicted_pd[\"prediction\"])"
+    "seaborn.histplot(predicted_pd[\"prediction\"], kde=True)\n"
    ]
   },
   {
@@ -334,7 +334,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "seaborn.distplot(predicted_pd[\"SalePrice\"] - predicted_pd[\"prediction\"])"
+    "seaborn.histplot(predicted_pd[\"SalePrice\"] - predicted_pd[\"prediction\"], kde=True)\n"
    ]
   },
   {
@@ -914,7 +914,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "seaborn.distplot(predicted_pd[\"label\"])"
+    "seaborn.histplot(predicted_pd[\"label\"], kde=True)\n"
    ]
   },
   {
@@ -923,7 +923,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "seaborn.distplot(predicted_pd[\"prediction\"])"
+    "seaborn.histplot(predicted_pd[\"prediction\"], kde=True)\n"
    ]
   },
   {
@@ -932,7 +932,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "seaborn.distplot(predicted_pd[\"label\"] - predicted_pd[\"prediction\"])"
+    "seaborn.histplot(predicted_pd[\"label\"] - predicted_pd[\"prediction\"], kde=True)\n"
    ]
   },
   {
@@ -965,4 +965,4 @@
  },
  "nbformat": 4,
  "nbformat_minor": 2
-}
+}

notebooks/exercises/spark-exercise-house-prices.ipynb

@@ -341,9 +341,9 @@
    "source": [
     "if ready:\n",
     "    predicted_pd = predicted[[\"SalePrice\", \"prediction\"]].toPandas()\n",
-    "    seaborn.distplot(predicted_pd[\"SalePrice\"])\n",
-    "    seaborn.distplot(predicted_pd[\"prediction\"])\n",
-    "    seaborn.distplot(predicted_pd[\"SalePrice\"] - predicted_pd[\"prediction\"])    "
+    "    seaborn.histplot(predicted_pd[\"SalePrice\"], kde=True)\n",
+    "    seaborn.histplot(predicted_pd[\"prediction\"], kde=True)\n",
+    "    seaborn.histplot(predicted_pd[\"SalePrice\"] - predicted_pd[\"prediction\"], kde=True)\n"
    ]
   },
   {
@@ -376,4 +376,4 @@
  },
  "nbformat": 4,
  "nbformat_minor": 2
-}
+}