Thalassohttp://mloss.orgUpdates and additions to ThalassoenMon, 22 Jul 2013 15:33:59 -0000Thalasso v0.2<html><p>Regularization paTH for LASSO problem (thalasso) </p> <p> thalasso solves problems of the following form: </p> <pre><code> minimize 1/2||X*beta-y||^2 + lambda*sum|beta_i|, </code></pre><p> where X and y are problem data and beta and lambda are variables. </p> <p>CALLING SEQUENCES [lambdas,betas,supports,alphas]=thalasso(X,y[,tolerance[,cond]]) </p> <p>INPUT X : NxP data matrix, N are the number of examples, P the number of features y : Nx1 data vector tolerance: scalar to indicate the last lambda accepted before 0 cond : scalar for matrix conditioning before inversion </p> <p>OUTPUT lambdas : 1xM vector containing lambdas of the regularization path betas : PxM matrix containing optimize beta vector for each lambda supports : 1xM cell containing non-null beta indexes for each lambda alphas : PxM matrix containing the sub-differential of each component for each lambda </p> <p>USAGE EXAMPLES [lambdas,betas]=thalasso(X,y); [lambdas,betas,supports,alphas]=thalasso(X,y,tolerance,cond) </p></html>Romain HERAULTMon, 22 Jul 2013 15:33:59 -0000