dual_ascent_fasta: Solve the projected dual ascent problem with fixed weights...

In wenshuoliu/ncvxclustr: None-Convex Clustering Algorithm

Description

Solve the projected dual ascent problem with fixed weights and adaptive step size and back-tracking

Usage

dual_ascent_fasta(X, Phi, weights, Lambda0, maxiter, eps, nv0, trace)

Arguments

X	the data, with the columns being units, the rows being features
Phi	the edge incidence matrix, defined as Phi_li = 1 if(l_1 == i); -1 if(l_2 == i); 0 otherwise
weights	the non-zero weights in a vector
Lambda0	the initial guess of Lambda
maxiter	maximum iterations
eps	the duality gap tolerence
trace	whether save the primal and dual values of every iteration
nv	initial step size

Value

a list including U, V, Lambda and number of iterations to convergence

wenshuoliu/ncvxclustr documentation built on May 4, 2019, 5:21 a.m.