matlab-(含教程)基于lq正则化的稀疏信号重建算法matlab仿真

function [xk,fcost,Bwhile,Time,mysnr]=FB_PPXALpLq(K,y,p,q,metric,alpha,beta,eta,xi,nbiter,xtrue) %%% This function defines the Trust region algorihtm based on %%% Forward-Backward algorithm %Initialization [~,N] = size(K); [xk_old, ~] = pds(K,y,xi,10); mysnr(1) = -10*log10(sum((xk_old-xtrue).^2) / sum(xtrue.^2)); fcost(1) = Fcost(xk_old,alpha,beta,eta,p,q); gamma = 1; prec = 1e-12; Bwhile = zeros(nbiter,1); fcost = zeros(nbiter,1); J = 5000; %ppxa max iterations %bmax = 10;% maximum TR iterates %metric 0: Lip constant, 1: FBVM without TR, 2: FBVM-TR L = ComputeLipschitz(alpha,beta,eta,p,q,N); %Algorithm for k=1:nbiter if (rem(k,100)==0) disp(['it=',num2str(k),': fcost = ',num2str(fcost(k-1))]) end tic; switch metric case 0 A = L*ones(N,1); B = A./gamma; xxk = xk_old - (1./B).*gradlplq(xk_old,alpha,beta,eta,p,q); xk = proxPPXAplus(K,B,xxk,y,xi,J,prec); case 1 A = condlplq(xk_old,alpha,beta,eta,p,q,0); B = A./gamma; xxk = xk_old - (1./B).*gradlplq(xk_old,alpha,beta,eta,p,q); xk = proxPPXAplus(K,B,xxk,y,xi,J,prec); case 2 ro = sum(abs(xk_old.^q))^(1/q); bwhile = 0; while 1 A = condlplq(xk_old,alpha,beta,eta,p,q,ro); B = A/gamma; xxk = xk_old - (1./B).*gradlplq(xk_old,alpha,beta,eta,p,q); xk = proxPPXAplus(K,B,xxk,y,xi,J,prec); if sum(abs(xk).^q)^(1/q) < ro ro = ro/2; bwhile = bwhile + 1; else break; end end Bwhile(k)= bwhile; end Time(k) = toc; error = norm(xk-xk_old)^2 / norm(xk_old)^2; mysnr(k+1) = -10*log10(sum((xk-xtrue).^2)/sum(xtrue.^2)); fcost(k+1) = Fcost(xk,alpha,beta,eta,p,q); if error < prec break; end xk_old = xk; end end