neingeist/coursera-ml-007-exercises
Archived
1
0
Fork 0
Code Releases Activity

Compute cost function for the neural network

Browse source
This commit is contained in:
neingeist 2014-11-01 20:30:58 +01:00
parent be6f3cbdef
commit f2154a8cc1
1 changed files with 24 additions and 6 deletions
Download patch file Download diff file Expand all files Collapse all files

30
ex4/nnCostFunction.m
View file

@ -8,8 +8,8 @@ function [J grad] = nnCostFunction(nn_params, ...
% [J grad] = NNCOSTFUNCTON(nn_params, hidden_layer_size, num_labels, ... % [J grad] = NNCOSTFUNCTON(nn_params, hidden_layer_size, num_labels, ...
% X, y, lambda) computes the cost and gradient of the neural network. The % X, y, lambda) computes the cost and gradient of the neural network. The
% parameters for the neural network are "unrolled" into the vector % parameters for the neural network are "unrolled" into the vector
% nn_params and need to be converted back into the weight matrices. % nn_params and need to be converted back into the weight matrices.
% %
% The returned parameter grad should be a "unrolled" vector of the % The returned parameter grad should be a "unrolled" vector of the
% partial derivatives of the neural network. % partial derivatives of the neural network.
% %
@ -24,8 +24,8 @@ Theta2 = reshape(nn_params((1 + (hidden_layer_size * (input_layer_size + 1))):en
% Setup some useful variables % Setup some useful variables
m = size(X, 1); m = size(X, 1);
% You need to return the following variables correctly % You need to return the following variables correctly
J = 0; J = 0;
Theta1_grad = zeros(size(Theta1)); Theta1_grad = zeros(size(Theta1));
Theta2_grad = zeros(size(Theta2)); Theta2_grad = zeros(size(Theta2));
@ -39,6 +39,24 @@ Theta2_grad = zeros(size(Theta2));
% cost function computation is correct by verifying the cost % cost function computation is correct by verifying the cost
% computed in ex4.m % computed in ex4.m
% %
X = [ones(m, 1), X]; % add a first colum of ones (bias term)
A_2 = sigmoid(X*Theta1');
A_2 = [ones(m, 1), A_2]; % (bias term)
A_3 = sigmoid(A_2*Theta2');
h_0 = A_3;
%disp(round(h_0));
% y is 1x5000 and holds the labels as numbers, turn it into 5000x10,
% each row holding the label as vectors, e.g. [0 1 0 0 0 ... ] for 2.
y = eye(num_labels)(y,:); % y is used as an index, it gets a row,
% e.g. [0 0 0 1 ... 0 0]
assert(size(y) == [m num_labels]);
J = 1/m * sum(sum(-y.*log(h_0) - (1-y).*log(1-h_0)));
assert(size(J) == [1 1]);
% Part 2: Implement the backpropagation algorithm to compute the gradients % Part 2: Implement the backpropagation algorithm to compute the gradients
% Theta1_grad and Theta2_grad. You should return the partial derivatives of % Theta1_grad and Theta2_grad. You should return the partial derivatives of
% the cost function with respect to Theta1 and Theta2 in Theta1_grad and % the cost function with respect to Theta1 and Theta2 in Theta1_grad and
@ -46,12 +64,12 @@ Theta2_grad = zeros(size(Theta2));
% that your implementation is correct by running checkNNGradients % that your implementation is correct by running checkNNGradients
% %
% Note: The vector y passed into the function is a vector of labels % Note: The vector y passed into the function is a vector of labels
% containing values from 1..K. You need to map this vector into a % containing values from 1..K. You need to map this vector into a
% binary vector of 1's and 0's to be used with the neural network % binary vector of 1's and 0's to be used with the neural network
% cost function. % cost function.
% %
% Hint: We recommend implementing backpropagation using a for-loop % Hint: We recommend implementing backpropagation using a for-loop
% over the training examples if you are implementing it for the % over the training examples if you are implementing it for the
% first time. % first time.
% %
% Part 3: Implement regularization with the cost function and gradients. % Part 3: Implement regularization with the cost function and gradients.