matlab-基于模糊小波神经网络FWNN的数据预测matlab仿真-源码

function [gradient]=Gradient_form(fmodel) [sample_num,feature_num]=size(fmodel.X); m=(fmodel.Pre)^feature_num; %pE_y=2/size(fmodel.X,1)*sum(fmodel.Y_est-fmodel.Y_true); one_mat=ones(feature_num,m); pE_y_inside=zeros(sample_num,1); pE_w=zeros(feature_num,m); pE_b=zeros(feature_num,m); pE_c=zeros(feature_num,m); pE_miu=zeros(pre,feature_num); pE_sigma=zeros(pre,feature_num); for i=1:sample_num pE_y_inside(i)=2*(fmodel.Y_est(i)-fmodel.Y_true(i)); x=fmodel.X(i,:); x=repmat(x',1,m); mid_w=((x-fmodel.B_mat)./fmodel.C_mat).^2; yita=repmat((fmodel.Yita(:,i))',feature_num,1); %E对w偏导(feature_num,m) pE_w=pE_w+pE_y_inside(i)*yita.*... ((one_mat-mid_w).*(exp(-1/2*mid_w))); %E对b偏导(feature_num,m) mid_b=((x-fmodel.B_mat)./fmodel.C_mat).^2; pE_b=pE_b+pE_y_inside(i)*yita.*fmodel.W_mat... .*((x-fmodel.B_mat)./(fmodel.C_mat.^2)).*... (3*one_mat-mid_b).*exp(-1/2*mid_b); %E对c偏导(feature_num,m) mid_c=((x-fmodel.B_mat).^2)./((fmodel.C_mat).^3); pE_c=pE_c+pE_y_inside(i)*yita.*fmodel.W_mat... .*mid_c.*(3*one_mat-mid_b).*exp(-1/2*mid_b); %y对A偏导(pre,feature_num) p_A_inside=Partial_a(fmodel.X(i,:),fmodel.Miu,... fmodel.Sigma,feature_num,pre,fmodel.Index,... fmodle.Phai(:,i),fmodel.Yita_all(i,1)); %E对miu偏导(pre,feature_num) mid_miu=((repmat(fmodel.X(i,:),pre,1)-fmodel.Miu)./... fmodel.Sigma).^2; pE_miu=pE_miu+pE_y_inside(i)*p_A_inside.*... (repmat(fmodel.X(i,:),pre,1)-fmodel.Miu)./... ((fmodel.Sigma.^2)).*exp(-1/2*mid_miu); %E对sigma偏导(pre,feature_num) pE_sigma=pE_sigma+pE_y_inside(i)*p_A_inside.*... ((repmat(fmodel.X(i,:),pre,1)-fmodel.Miu).^2)./... ((fmodel.Sigma.^3)).*exp(-1/2*mid_miu); end pE_w=pE_w/sample_num; pE_b=pE_b/sample_num; pE_c=pE_c/sample_num; pE_miu=pE_miu/sample_num; pE_sigma=pE_sigma/sample_num; % gradient=struct('pE_w',pE_w,'pE_b',pE_b,'pE_c',pE_c... % ,'pE_miu',pE_miu,'pE_sigma',pE_sigma); gradient=[pE_miu(:);pE_sigma(:);pE_b(:);pE_b(:);pE_w(:)]; end