PCA

crank_ml.decomposition.pca.PCA

Performs linear Principal Component Analysis (PCA) on a low-rank matrix, batches of such matrices, or sparse matrix. This approach uses the nearly optimal approximation of a singular value decomposition of a centered matrix.

Our implementation is a wrapper over the pytorch implementation with polyak averaging over the weights. As a consequence this may not be stable.

Parameters

Parameter	Description
n_features	Number of input features
n_components	(Default: 6) Maximum number of components to keep.
polyak_weight	(Default: 0.1) The weight update rate, a reasonable value if between 0 and 1

Example

import numpy as np
import torch

from crank_ml.decomposition.pca import PCA

X = torch.from_numpy(np.array([[1, 2, 3], [1, 4, 5], [1, 0, 1], [10, 2, 2], [10, 4, 3], [10, 0, 1]])).float()

pca = PCA(n_features=3, n_components=2)

for _ in range(5):
    _ = pca(X)

pca.eval()
pca_decomposition = pca(X)