11 Mapping to lower dimensions
11.1 Manifold learning
Isomap
Local linear embedding
Multi dimensional scaling
11.2 Decomposition techniques
Singular value decomposition
Singular value decomposition is used to compress large matrices of your data into smaller ones, with much less data, but without loosing a lot of information. Please visit the mathematical explanation for the underlying mechanisms.
from sklearn.decomposition import TruncatedSVD
svd = TruncatedSVD(n_components=10)
svd.fit_transform(X_train)
svd.transform(X_test)