RWSRepLearning.m

% This script generates the feature representation of each time series by 
% computing random features between random series and raw time-series. We
% use dynamic time warping to compute the distance between a pair of
% time-series. 
%
% Author: Lingfei Wu
% Date: 01/20/2019

function ZRep = RWSRepLearning(DS,sigma,R,DMin,DMax)
    
    % load data and generate corresponding train and test data
   

    trainX = DS.Train;
    %trainy = trainData(:,1);
    testX = DS.Test;
    %testy = testData(:,1);    
    
    


    % generate random time series with variable length, where each value in
    % random series is sampled from Gaussian distribution parameterized by sigma. 
    timer_start = tic;
    rng('default')
    sampleX = cell(R,1);
    for i=1:R
        D = randi([DMin, DMax],1);
        sampleX{i} = randn(1, D)./sigma; % gaussian
    end
    [trainFeaX_random, train_dtw_time] = dtw_similarity_cell(trainX, sampleX);
    trainFeaX_random = trainFeaX_random/sqrt(R); 
    [testFeaX_random, test_dtw_time] = dtw_similarity_cell(testX, sampleX);
    testFeaX_random = testFeaX_random/sqrt(R); 
    Train = trainFeaX_random;
    Test = testFeaX_random;
    
    ZRep = [Train;Test];
    
    %telapsed_random_fea_gen = toc(timer_start);
    
    % Note: real_total_end_time is the real total time, including both dtw
    % and ground distance, of generating both train and test features using 
    % multithreads. user_dtw_time is the real time that accounts for 
    % computation of dtw with one thread. 
    %Runtime.real_total_dtw_time = telapsed_random_fea_gen;
    %Runtime.user_dtw_time = train_dtw_time + test_dtw_time;
    %Runtime.user_train_dtw_time = train_dtw_time;
    %Runtime.user_test_dtw_time = test_dtw_time;
end