Evaluating the Model Performance

Use show_reconstructions function and pass stacked_ae and X_test as input argument. This displays 5 ground truth images and the corresponding reconstructed images.

show_reconstructions(stacked_ae, X_test)

Let us view the rounded_accuracies of X_test using stacked_ae.evaluate.

<< your code comes here >>(X_test, X_test)

Let us view the class-wise clusters for the validation data as predicted by our model stacked_ae. Since we can't display multiple-dimensions, we shall do this by using TSNE dimensionality reduction.

• We shall use predict of stacked_encoder on X_valid to get the compressed data of the validation data.

• Use fit_transform of TSNE() to get the 2D representation of the compressed validation data and scale it data.

• Now plot this data with colormaps for each class.

Use the following code to do get the 2D representation of the compressed validation data.

np.random.seed(42)

from sklearn.manifold import TSNE

start = time.time()

X_valid_compressed = stacked_encoder.predict(X_valid)
tsne = TSNE()
X_valid_2D = tsne.fit_transform(X_valid_compressed)
X_valid_2D = (X_valid_2D - X_valid_2D.min()) / (X_valid_2D.max() - X_valid_2D.min())

end = time.time()

print("Time of execution:", round(end-start,2),"seconds")

Use the following code to display the class-wise clusters.

plt.figure(figsize=(10, 8))
cmap = plt.cm.tab10
plt.scatter(X_valid_2D[:, 0], X_valid_2D[:, 1], c=y_valid, s=10, cmap=cmap)
image_positions = np.array([[1., 1.]])
for index, position in enumerate(X_valid_2D):
    dist = np.sum((position - image_positions) ** 2, axis=1)
    if np.min(dist) > 0.02: # if far enough from other images
        image_positions = np.r_[image_positions, [position]]
        imagebox = mpl.offsetbox.AnnotationBbox(
            mpl.offsetbox.OffsetImage(X_valid[index], cmap="binary"),
            position, bboxprops={"edgecolor": cmap(y_valid[index]), "lw": 2})
        plt.gca().add_artist(imagebox)
plt.axis("off")
plt.show()