demoPCA.m

input_data = rand(1000,3);

%% Visualize Input Data
figure(1);
hold off;
plot3(input_data(:,1), input_data(:,2), input_data(:,3), 'ro');

%% Function PCA, input_data, out_dim
% Use this to switch methods
use_svd_method=1;

out_dim = 2;
[count in_dim] = size(input_data);

% SVD
if (use_svd_method)
    sub_input_data = (input_data - repmat(mean(input_data),count,1))/sqrt(count-1);
    [U,S,V] = svd(sub_input_data);
    % First out_dim columns as PCA bases
    pcaV = V(:,1:out_dim);
    output_data = input_data * pcaV;
else
    % EIG
    mean_input_data = mean(input_data);
    sub_input_data = input_data - repmat(mean_input_data, count,1);
    mean_mat = sub_input_data' * sub_input_data ./ (count - 1);
    cov_mat = mean_mat;
    [V D] = eig(cov_mat);
    % Last out_dim columns as PCA bases
    pcaV = V(:,in_dim - out_dim + 1: in_dim);
    output_data = input_data * pcaV;
end

%% End Function PCA

%% Visualize Output Data
figure(2);
hold off;
plot(output_data(:,1), output_data(:,2), 'bo');