def _mixed_norm_solver_cd(M,
G,
alpha,
lipschitz_constant,
maxit=10000,
tol=1e-8,
verbose=None,
init=None,
n_orient=1):
"""Solves L21 inverse problem with coordinate descent"""
from sklearn.linear_model.coordinate_descent import MultiTaskLasso
n_sensors, n_times = M.shape
n_sensors, n_sources = G.shape
if init is not None:
init = init.T
clf = MultiTaskLasso(alpha=alpha / len(M),
tol=tol,
normalize=False,
fit_intercept=False,
max_iter=maxit,
warm_start=True)
clf.coef_ = init
clf.fit(G, M)
X = clf.coef_.T
active_set = np.any(X, axis=1)
X = X[active_set]
gap, pobj, dobj, _ = dgap_l21(M, G, X, active_set, alpha, n_orient)
return X, active_set, pobj
class MultiTaskLassoImpl():
def __init__(self,
alpha=1.0,
fit_intercept=True,
normalize=False,
copy_X=True,
max_iter=1000,
tol=0.0001,
warm_start=False,
random_state=None,
selection='cyclic'):
self._hyperparams = {
'alpha': alpha,
'fit_intercept': fit_intercept,
'normalize': normalize,
'copy_X': copy_X,
'max_iter': max_iter,
'tol': tol,
'warm_start': warm_start,
'random_state': random_state,
'selection': selection
}
def fit(self, X, y=None):
self._sklearn_model = SKLModel(**self._hyperparams)
if (y is not None):
self._sklearn_model.fit(X, y)
else:
self._sklearn_model.fit(X)
return self
def predict(self, X):
return self._sklearn_model.predict(X)
def _mixed_norm_solver_cd(M,
G,
alpha,
lipschitz_constant,
maxit=10000,
tol=1e-8,
verbose=None,
init=None,
n_orient=1,
dgap_freq=10):
"""Solve L21 inverse problem with coordinate descent."""
from sklearn.linear_model.coordinate_descent import MultiTaskLasso
assert M.ndim == G.ndim and M.shape[0] == G.shape[0]
clf = MultiTaskLasso(alpha=alpha / len(M),
tol=tol / sum_squared(M),
normalize=False,
fit_intercept=False,
max_iter=maxit,
warm_start=True)
if init is not None:
clf.coef_ = init.T
else:
clf.coef_ = np.zeros((G.shape[1], M.shape[1])).T
clf.fit(G, M)
X = clf.coef_.T
active_set = np.any(X, axis=1)
X = X[active_set]
gap, p_obj, d_obj, _ = dgap_l21(M, G, X, active_set, alpha, n_orient)
return X, active_set, p_obj
def _mixed_norm_solver_cd(M, G, alpha, maxit=10000, tol=1e-8, verbose=None, init=None, n_orient=1):
"""Solves L21 inverse problem with coordinate descent"""
from sklearn.linear_model.coordinate_descent import MultiTaskLasso
n_sensors, n_times = M.shape
n_sensors, n_sources = G.shape
if init is not None:
init = init.T
clf = MultiTaskLasso(
alpha=alpha / len(M), tol=tol, normalize=False, fit_intercept=False, max_iter=maxit, warm_start=True
)
clf.coef_ = init
clf.fit(G, M)
X = clf.coef_.T
active_set = np.any(X, axis=1)
X = X[active_set]
gap, pobj, dobj, _ = dgap_l21(M, G, X, active_set, alpha, n_orient)
return X, active_set, pobj
def _mixed_norm_solver_cd(M, G, alpha, lipschitz_constant, maxit=10000,
tol=1e-8, verbose=None, init=None, n_orient=1,
dgap_freq=10):
"""Solve L21 inverse problem with coordinate descent."""
from sklearn.linear_model.coordinate_descent import MultiTaskLasso
assert M.ndim == G.ndim and M.shape[0] == G.shape[0]
clf = MultiTaskLasso(alpha=alpha / len(M), tol=tol / sum_squared(M),
normalize=False, fit_intercept=False, max_iter=maxit,
warm_start=True)
if init is not None:
clf.coef_ = init.T
else:
clf.coef_ = np.zeros((G.shape[1], M.shape[1])).T
clf.fit(G, M)
X = clf.coef_.T
active_set = np.any(X, axis=1)
X = X[active_set]
gap, p_obj, d_obj, _ = dgap_l21(M, G, X, active_set, alpha, n_orient)
return X, active_set, p_obj