Created using Colaboratory

PyTorch_Explore/1_Regression.ipynb

@@ -5,7 +5,7 @@
     "colab": {
       "provenance": [],
       "toc_visible": true,
-      "authorship_tag": "ABX9TyPcflmMmcoKWMeJCvnHkpzn",
+      "authorship_tag": "ABX9TyMuXeJk+Zh3Dm2wwP6dBkFI",
       "include_colab_link": true
     },
     "kernelspec": {
@@ -72,15 +72,16 @@
       "cell_type": "code",
       "source": [
         "# 0) Prepare Data\n",
-        "X_numpy, y_numpy = datasets.make_regression(n_samples=100,\n",
-        "                                            n_features=1,\n",
-        "                                            noise=20,\n",
+        "X_numpy, y_numpy = datasets.make_regression(n_samples=100, # Number of datapoints\n",
+        "                                            n_features=1, # Number of features\n",
+        "                                            noise=20, # random noise added to the target values\n",
         "                                            random_state=1)\n",
-        "X = torch.from_numpy(X_numpy.astype(np.float32))\n",
-        "y = torch.from_numpy(y_numpy.astype(np.float32))\n",
-        "y = y.view(y.shape[0], 1)\n",
+        "X = torch.from_numpy(X_numpy.astype(np.float32)) # Converts NumPy array to a PyTorch tensor\n",
+        "y = torch.from_numpy(y_numpy.astype(np.float32)) # Converts NumPy array to a PyTorch tensor\n",
+        "y = y.view(y.shape[0], 1) # Reshape the target tensor to have a shape of (100, 1)\n",
+        "# the target tensor should have the same number of columns as the input tensor\n",
         "\n",
-        "n_samples, n_features = X.shape"
+        "n_samples, n_features = X.shape # store the number of samples and features in the data"
       ],
       "metadata": {
         "id": "OWVhelS0UreC"
@@ -92,9 +93,12 @@
       "cell_type": "code",
       "source": [
         "# 1) model\n",
+        "\n",
+        "# store the number of input and output features for the model\n",
         "input_size = n_features\n",
         "output_size = 1\n",
-        "model = nn.Linear(input_size, output_size)"
+        "\n",
+        "model = nn.Linear(input_size, output_size) # Linead Regression Model"
       ],
       "metadata": {
         "id": "ODgAKQuPUc1b"
@@ -106,8 +110,8 @@
       "cell_type": "code",
       "source": [
         "# 2) loss and optimizer\n",
-        "learning_rate = 0.01\n",
-        "criterion = nn.MSELoss()\n",
+        "learning_rate = 0.01 # specify the learning rate of the optimizer\n",
+        "criterion = nn.MSELoss() # loss function\n",
         "optimizer = torch.optim.SGD(model.parameters(), lr = learning_rate)"
       ],
       "metadata": {
@@ -125,15 +129,15 @@
         "for epoch in range(num_epochs):\n",
         "    # forward pass and loss\n",
         "    y_predicted = model(X)\n",
-        "    loss = criterion(y_predicted, y)\n",
+        "    loss = criterion(y_predicted, y) # calculate the loss between the predictions and the ground truth labels\n",
         "\n",
         "    # backward pass\n",
-        "    loss.backward()\n",
+        "    loss.backward() # backpropagates the loss through the model\n",
         "\n",
         "    # update weights\n",
         "    optimizer.step()\n",
         "\n",
-        "    optimizer.zero_grad()\n",
+        "    optimizer.zero_grad() # sets the gradients of the model to zero\n",
         "\n",
         "    if (epoch+1) % 10 == 0:\n",
         "        print(f'epoch: {epoch+1}, loss = {loss.item():.4f}')"
@@ -169,7 +173,7 @@
       "cell_type": "code",
       "source": [
         "# Plot\n",
-        "predicted = model(X).detach().numpy()\n",
+        "predicted = model(X).detach().numpy() # converts prediction to numpy arrays\n",
         "plt.plot(X_numpy, y_numpy, 'ro')\n",
         "plt.plot(X_numpy, predicted, 'b')\n",
         "plt.show()"