PCA with SVD using pytorch

pytorch_pca_svd.py

def torch_pca_svd(X, center=True):
    n, _ = X.shape
    # center points along axes
    if center:
        X = X - X.mean(dim=0)
    # perform singular value decomposition
    u, s, v = torch.svd(X)
    # extract components
    components = v.T
    explained_variance = torch.mul(s, s) / (n-1)
    return components, explained_variance