% addpath('C:\Users\d035190\Documents\MATLAB\Toolboxes\Deep Neural Network\DeepNeuralNetwork')

% data_mean = mean(norm_temp,1);
% data_std = std(norm_temp);
% numcases=length(norm_temp);
% norm_data =( norm_temp - repmat(data_mean,numcases,1) ) ./ repmat( data_std, numcases,1);
% plot(norm_data)
% 
% data_mean = mean(oraInput,1);
% data_std = std(oraInput);
% numcases=length(oraInput);
% norm_ora =( oraInput - repmat(data_mean,numcases,1) ) ./ repmat( data_std, numcases,1);
% 
% 
% data_mean = mean(meseInput,1);
% data_std = std(meseInput);
% numcases=length(meseInput);
% norm_mese =( meseInput - repmat(data_mean,numcases,1) ) ./ repmat( data_std, numcases,1);
% 
% data_mean = mean(Day,1);
% data_std = std(Day);
% numcases=length(Day);
% norm_day =( Day- repmat(data_mean,numcases,1) ) ./ repmat( data_std, numcases,1);
% 
% zs=zscore(filteredtemp);

load('lowpassfilteredpreprocesseddatafordeepnn17.mat')

    inputtrain= inputdata(1:120000,:);
    inputtest= inputdata(120000:end,:);


    outputtrain= outputrain3hour(1:120000,1);
    outputtest=  outputrain3hour(120000:end,1);



    num = 120000;
    nodes = [12 1000 600 400 100 1];

    % input=[norm_date norm_time]
    % input=[norm_date norm_time]

    dnn1 = randDBN( nodes,'BBPDBN');


    IN=inputtrain;
    OUT=outputtrain;

    % IN = [inputdata(3:70000,:) norm_temp(2:69999) norm_temp(1:69998)];
    % OUT = norm_temp(3:70000);

    %dnn = randDBN( nodes, 'BBPDBN' ); % ICPR 2014
    %dnn = randDBN( nodes, 'GBDBN' );
    nrbm = numel(dnn1.rbm);

    opts.MaxIter = 1;
    opts.BatchSize = 10;
    opts.Verbose = true;
    opts.StepRatio =0.1;
    opts.Layer = nrbm-1;
    opts.DropOutRate = 0.0;
    opts.Object = 'Square';

    dnn1 = pretrainDBN(dnn1, IN, opts);
    estimate=v2h(dnn1,IN);
    figure
    plot(estimate);

    dnn1= SetLinearMapping(dnn1, IN, OUT);
    estimate1=v2h(dnn1,IN);
    figure
    plot(estimate1);

    %dnnm=dnn1
    % dnn1=dnnm

    opts.Layer = 0;
    opts.MaxIter=1000;
    opts.StepRatio =1;
    dnn1 = trainDBN(dnn1, IN, OUT, opts);
    estimate2=v2h(dnn1,IN);
    figure
    plot(estimate2);
    rmse = CalcRmse(dnn1, IN, OUT);
    rmse
    figure
    
    hold on 
    plot(OUT)
    hold on
    plot(estimate2)
    
    

% opts.Layer = 0;
%     opts.MaxIter=10;
%     opts.StepRatio =1;
%     dnn1 = trainDBN(dnn1, IN, OUT, opts);
%     estimate2=v2h(dnn1,IN);
%     figure
%     plot(estimate2);
%     rmse = CalcRmse(dnn1, IN, OUT);
%     rmse
%       
%     
%     opts.Layer = 0;
%     opts.MaxIter=10;
%     opts.StepRatio =1;
%     dnn1 = trainDBN(dnn1, IN, OUT, opts);
%     estimate2=v2h(dnn1,IN);
%     figure
%     plot(estimate2);
%     rmse = CalcRmse(dnn1, IN, OUT);
%     rmse

estimatetest=v2h(dnn1,inputtest);
rmsetest = CalcRmse(dnn1, inputtest, outputtest);

save deepnnpredictionlowpas17.mat