sparse_ho.algo.Backward

class sparse_ho.algo.Backward(use_stop_crit=True, verbose=False)

Algorithm to compute the hypergradient using backward differentiation.

The algorithm first computes the regression coefficients beta, using proximal coordinate descent, storing all the iterates. Then the gradient is computed in a backward way.

Parameters
use_stop_crit: bool, optional (default=True)

Use stopping criterion in hypergradient computation. If False, run to maximum number of iterations.

verbose: bool, optional (default=False)

Verbosity of the algorithm.

__init__(use_stop_crit=True, verbose=False)

Initialize self. See help(type(self)) for accurate signature.

Methods

__init__([use_stop_crit, verbose])

Initialize self.

compute_beta_grad(X, y, log_alpha, model, …)

Compute beta and hypergradient with backward differentiation of proximal coordinate descent.