Mastering Machine Learning Algorithms

Example of t-distributed stochastic neighbor embedding 

We can apply this powerful algorithm to the same Olivetti faces dataset, using the Scikit-Learn class TSNE with n_components=2 and perplexity=20:

from sklearn.manifold import TSNEtsne = TSNE(n_components=2, perplexity=20)X_tsne = tsne.fit_transform(faces['data'])

The result for all 400 samples is shown in the following graph:

t-SNE applied to the Olivetti faces dataset

A visual inspection of the label distribution can confirm that t-SNE recreated the optimal clustering starting from the original high-dimensional distribution. This algorithm can be employed in several non-linear dimensionality ...