function [] = nn()

data = load('train_data.mat');

train_x = data.x(2:end, :);

train_y = data.t;

losses = []; train_errors = []; test_wrongs = [];

%Here we perform mini-batch stochastic gradient descent %If batchsize = 1, it would be stochastic gradient descent %If batchsize = N, it would be basic gradient descent

batchsize = 50;

%Num of batches

numbatches = size(train_x, 2) / batchsize;

%% Training part %Learning rate alpha

alpha = 0.01;

%Lambda is for regularization

lambda = 0.001;

%Num of iterations

numepochs = 20;

numOfHiddenLayer = 2;

s{1} = size(train_x, 1); s{2} = 200; s{3} = 10; s{4} = 100; s{5} = 10;

for i = 1 : numOfHiddenLayer

W{i} = zeros(s{i+1}, s{i});

b{i} = 0;

end

for j = 1 : numepochs %randomly rearrange the training data for each epoch %We keep the shuffled index in kk, so that the input and output could %be matched together

kk = randperm(size(train_x, 2));

for l = 1 : numbatches

%Set the activation of the first layer to be the training data %while the target is training labels

a{1} = train_x(:, kk((l-1)*batchsize+1 : l*batchsize ));

y = train_y(:, kk((l-1)*batchsize+1 : l*batchsize ));

%Forward propagation, layer by layer %Here we use sigmoid function as an example

for i = 2 : numOfHiddenLayer + 1

a{i} = sigm( bsxfun(@plus, W{i-1}*a{i-1}, b{i-1}) );

end

%Calculate the error and back-propagate error layer by layers

d{numOfHiddenLayer + 1} = -(y - a{numOfHiddenLayer + 1}) .* a{numOfHiddenLayer + 1} .* (1-a{numOfHiddenLayer + 1});

for i = numOfHiddenLayer : -1 : 2

d{i} = W{i}' * d{i+1} .* a{i} .* (1-a{i});

end %Calculate the gradients we need to update the parameters %L2 regularization is used for W

for i = 1 : numOfHiddenLayer

dW{i} = d{i+1} * a{i}';

db{i} = sum(d{i+1}, 2);

W{i} = W{i} - alpha * (dW{i} + lambda * W{i});

b{i} = b{i} - alpha * db{i};

end

end

end

[~, cr] = NN_classify(train_x, W, b, numOfHiddenLayer);

[acc, ~, ~] = accuracy(data.labels, cr);

acc

end

function [acc, correct, wrong] = accuracy(labels, class_result)

count = 0;

total_count=length(labels);

for i=1:length(labels)

if labels(i)==class_result(i)

count = count+1;

end

end

correct = count;

wrong = total_count-correct;

acc = correct*100/total_count;

end

function [res, class_result] = NN_classify(x, W, b, numOfHiddenLayer)

a{1} = x;

for i = 2 : numOfHiddenLayer + 1

a{i} = sigm( bsxfun(@plus, W{i-1}*a{i-1}, b{i-1}) );

end

res = 0;

[~, class_result] = max(a{numOfHiddenLayer + 1});

class_result = mod(class_result, 10);

save('temp_res', 'class_result');

end

function [yn] = softmax_activation(an)

[a,b] = size(an);

yn = zeros(a,b);

denom = sum(exp(an));

for i=1:b

yn(:,i) = exp(an(:,i))./denom(i);

end

end

function [res] = h(bn)

res = sigm(bn);

%res = relu(bn);

end