matlab gpuarray 非常慢,我使用gpuArray进行矩阵运算发现消耗时间要大于cpu下

程序代码: [选择]function [cost grad] = SparseEncoder( netConfig, data, weight, predOnly )

%% this is a basic autoencoder only weight sparse penalize

if ~exist('predOnly', 'var')

predOnly = false;

end

[n, m] = size(data);

tic();

netConfig = stack2param(netConfig, weight);

toc();

w1 = netConfig.layers{1}.w;

b1 = netConfig.layers{1}.b;

w2 = netConfig.layers{2}.w;

b2 = netConfig.layers{2}.b;

% feedforward

z2 = bsxfun(@plus, w1*data, b1);

a2 = sigmoid(z2);

z3 = bsxfun(@plus, w2*a2, b2);

a3 = sigmoid(z3);

% compute kl distance

rouBar = sum(a2, 2)./m;

rou = netConfig.sparsity;

klPenatyItem = rou*log(rou./rouBar)+(1-rou)*log((1-rou)./(1-rouBar));

%cost

cost = 1/(2*m)*(sum