Nonlinear dimensionality reduction and visualization
In the previous section, we covered linear transformation techniques, such as PCA and LDA, for feature extraction. In this section, we will discuss why considering nonlinear dimensionality reduction techniques might be worthwhile.
One nonlinear dimensionality reduction technique that is particularly worth highlighting is t-distributed stochastic neighbor embedding (t-SNE) since it is frequently used in literature to visualize high-dimensional datasets in two or three dimensions. We will see how we can apply t-SNE to plot images of handwritten images in a 2-dimensional feature space.
Why consider nonlinear dimensionality reduction?
Many machine learning algorithms make assumptions about the linear separability of the input data. You have learned that the perceptron even requires perfectly linearly separable training data to converge. Other algorithms that we have covered so far assume that the lack of perfect linear separability...
