-
Notifications
You must be signed in to change notification settings - Fork 4
/
softmaxCost.m
38 lines (27 loc) · 1.14 KB
/
softmaxCost.m
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
function [cost, grad] = softmaxCost(theta, numClasses, inputSize, lambda, data, labels)
% numClasses - the number of classes
% inputSize - the size N of the input vector
% lambda - weight decay parameter
% data - the N x M input matrix, where each column data(:, i) corresponds to
% a single test set
% labels - an M x 1 matrix containing the labels corresponding for the input data
%
% Unroll the parameters from theta
theta = reshape(theta, numClasses, inputSize);
numCases = size(data, 2); %num of training examples
groundTruth = full(sparse(labels, 1:numCases, 1)); % groundTruth(i,j) = 1 if y(j) = i else o
cost = 0;
thetagrad = zeros(numClasses, inputSize);
M = theta*data;
M = bsxfun(@minus, M, max(M, [], 1));
M = exp(M);
M = bsxfun(@rdivide, M, sum(M));
S = groundTruth - M;
for j = 1:numClasses
thetagrad(j,:) = lambda*theta(j,:) - numCases.\transpose(sum(data.*repmat(S(j,:),inputSize,1),2));
end
cost = 0.5*lambda*sum(sum(theta.*theta)) - sum(sum(log(M).*groundTruth))/numCases;
% ------------------------------------------------------------------
% Unroll the gradient matrices into a vector for minFunc
grad = [thetagrad(:)];
end