himalaya.utils.compute_lipschitz_constants¶

himalaya.utils.compute_lipschitz_constants(Xs, kernelize='XTX', random_state=None)[source]¶

Compute Lipschitz constants of gradients of linear regression problems.

Find the largest eigenvalue of X^TX for several X, using power iteration.

Parameters

Xsarray of shape (n_kernels, n_samples, n_features) or (n_kernels, n_samples, n_samples): Multiple linear features or kernels.
kernelizestr in {“XTX”, “XXT”, “X”}: Whether to consider X^TX, XX^T, or directly X.
random_stateint, or None: Random generator seed. Use an int for deterministic search.

Returns