iterreg.sparse.estimators.SparseIterReg

class iterreg.sparse.estimators.SparseIterReg(train_ratio=0.8, f_test=1, max_iter=1000, memory=20, prox=CPUDispatcher(<function shrink>), step_ratio=1, verbose=False)

__init__(train_ratio=0.8, f_test=1, max_iter=1000, memory=20, prox=CPUDispatcher(<function shrink>), step_ratio=1, verbose=False)

Sparse Recovery with iterative regularization. Chambolle Pock iterations are performed on min J(w) s.t. Xw = y as long as the test MSE decreases.

Parameters

train_ratiofloat, optional (default=0.8)

Fraction of the samples used in the training set.

f_testint, optional (default=1)

The criterion to stop the solver is tested every f_test iterations.

max_iterint, optional (default=1000)

Maximum number of iterations performed.

memoryint, optional (default=20)

If the criterion does not decrease for memory computation, the solver stops.

prox: callable

The proximal operator of the regularizer J at level tau. By default, shrink is used, corresponding to L1.

step_ratiofloat, optional (default=1)

Ratio between primal and dual stepsizes of the algorithm. A higher step may slow down convergence, but improve the sparsity of the best iterate.

verbose: bool, optional (default=False)

Solver verbosity.

Methods

__init__([train_ratio, f_test, max_iter, ...])	Sparse Recovery with iterative regularization.
debias(X, y)
fit(X, y[, train_idx, test_idx])	Fit model.
get_params([deep])	Get parameters for this estimator.
predict(X)	Predict using the linear model.
set_params(**params)	Set the parameters of this estimator.