graphtoolbox.utils.GL_3SR

Functions

f_obj(Y, X, H, lbd, alpha, beta[, gamma])	function objectif from paper
get_X_complete_graph(p)	get X for a complete graph

Classes

FGL_3SR([beta, alpha, trace, gamma, X0, ...])

class graphtoolbox.utils.GL_3SR.FGL_3SR(beta=0.1, alpha=0.1, trace=1, gamma=0.0, X0=None, lbd0=None, H0=None, cv_crit=0.01, maxit=1000, keep_best=True, verbose=True)

Bases: object

gamma

Values on constraints

trace

Initial guess

H0

Values on variables of the algorithms

cv_crit

Variables of interest

lbd

Error sequence

fit(Y, norm='l1')

fit all the variables (X, H, Lambda)

f_obj(Y)

obj fct with current values H, X, Lambda

grad_f_X(Y)

gradient of the obj fct of the problem in X with current values H

prepare_fit(Y)

initialisation of all the variables

solve_H0_closedform(Y, p, n)

H update with a closeform when l0_2 norm is choosed

solve_H0_closedform_K(Y, p, n)

H update with a closeform when l0_2 norm is choosed ( < K)

solve_H1_closedform(Y, p, n)

H update with a closeform when l1_2 norm is choosed

solve_lbd(I_p)

lbd=diag(Lambda) update with linear programming

solve_X_relaxation(Y, p)

X update with the relaxation method assuming an unique connex component

get_coeffs(order=True)

return X, H, lbd and the error trajectory

prepare_constraints(alpha, X, H)

prepare variables for the constraints funtion

compute_constraints_Diag(alpha=None, X=None, H=None, trace=None)

compute the constraints for the linear (quadratic) program in Lambda starting point: complete graph

graphtoolbox.utils.GL_3SR.f_obj(Y, X, H, lbd, alpha, beta, gamma=0.0)

function objectif from paper

graphtoolbox.utils.GL_3SR.get_X_complete_graph(p)

get X for a complete graph