paths¶
Module: paths
¶
Inheritance diagram for regreg.paths
:
Classes¶
lasso
¶
-
class
regreg.paths.
lasso
(loss_factory, X, penalty_structure=None, group_weights={}, elastic_net=identity_quadratic(0.000000, 0.0, 0.0, 0.000000), alpha=0.0, intercept=True, positive_part=None, unpenalized=None, lagrange_proportion=0.05, nstep=100, scale=True, center=True)¶ Bases:
object
-
__init__
(loss_factory, X, penalty_structure=None, group_weights={}, elastic_net=identity_quadratic(0.000000, 0.0, 0.0, 0.000000), alpha=0.0, intercept=True, positive_part=None, unpenalized=None, lagrange_proportion=0.05, nstep=100, scale=True, center=True)¶ Initialize self. See help(type(self)) for accurate signature.
-
property
Xn
¶
-
property
active
¶
-
construct_loss
(candidate_set, lagrange)¶
-
property
elastic_net
¶
-
get_lagrange
()¶
-
get_lagrange_sequence
()¶
-
grad
(loss=None)¶ Gradient at current value. This includes the gradient of the smooth loss as well as the gradient of the elastic net part. This is used for determining whether the KKT conditions are met and which coefficients are in the strong set.
-
property
lagrange
¶
-
property
lagrange_max
¶
-
property
lagrange_sequence
¶
-
property
lipschitz
¶
-
classmethod
logistic
(X, Y, *args, **keyword_args)¶
-
property
loss
¶
-
main
(inner_tol=1e-05, verbose=False)¶
-
property
nonzero
¶
-
property
null_solution
¶
-
property
problem
¶
-
restricted_problem
(candidate_set, lagrange)¶ Assumes the candidate set includes intercept as first column.
-
set_lagrange
(lagrange)¶
-
set_lagrange_sequence
(lagrange_sequence)¶
-
property
shape
¶
-
slice_columns
(columns)¶
-
property
solution
¶
-
solve_subproblem
(candidate_set, lagrange_new, **solve_args)¶
-
classmethod
squared_error
(X, Y, *args, **keyword_args)¶
-
strong_set
(lagrange_cur, lagrange_new, grad=None, slope_estimate=1)¶
-
nesta
¶
-
class
regreg.paths.
nesta
(loss_factory, X, atom_factory, epsilon=None, **lasso_keywords)¶ Bases:
regreg.paths.lasso
-
__init__
(loss_factory, X, atom_factory, epsilon=None, **lasso_keywords)¶ Initialize self. See help(type(self)) for accurate signature.
-
property
Xn
¶
-
property
active
¶
-
construct_loss
(candidate_set, lagrange)¶
-
property
elastic_net
¶
-
property
epsilon
¶
-
property
final_step
¶
-
get_dual_term
(lagrange)¶
-
get_epsilon
()¶
-
get_final_step
()¶
-
get_lagrange
()¶
-
get_lagrange_sequence
()¶
-
grad
(loss=None)¶ Gradient at current value. This includes the gradient of the smooth loss as well as the gradient of the elastic net part. This is used for determining whether the KKT conditions are met and which coefficients are in the strong set.
-
property
lagrange
¶
-
property
lagrange_max
¶
-
property
lagrange_sequence
¶
-
property
lipschitz
¶
-
classmethod
logistic
(X, Y, *args, **keyword_args)¶
-
property
loss
¶
-
main
(inner_tol=1e-05, verbose=False)¶
-
property
nonzero
¶
-
property
null_solution
¶
-
property
problem
¶
-
restricted_problem
(candidate_set, lagrange)¶ Assumes the candidate set includes intercept as first column.
-
set_dual_term
(lagrange, dual_term)¶
-
set_epsilon
(epsilon)¶
-
set_final_step
(value)¶
-
set_lagrange
(lagrange)¶
-
set_lagrange_sequence
(lagrange_sequence)¶
-
property
shape
¶
-
slice_columns
(columns)¶
-
property
solution
¶
-
solve_subproblem
(candidate_set, lagrange_new, **solve_args)¶
-
classmethod
squared_error
(X, Y, *args, **keyword_args)¶
-
strong_set
(lagrange_cur, lagrange_new, grad=None, slope_estimate=1)¶
-