xujingxu / pgbm-rbm Goto Github PK

This matlab package contains demo codes of point-wise Gated Boltzmann 
machine on variations of MNIST database (mnist-back-image, mnist-back-rand)
running on the MATLAB with or without GPU Jacket.


Instructions:

a1. download and unarchive the code package.

a2. download MNIST variations datasets (mnist-back-image, mnist-back-rand) 
    from the url below and place them into data folder:
    
    http://www.iro.umontreal.ca/~lisa/twiki/bin/view.cgi/Public/MnistVariations

a3. download and compile the latest liblinear code package (liblinear-1.93)
    from the following url:

    http://www.csie.ntu.edu.tw/~cjlin/liblinear/

b.  open MATLAB and provide following information.

    1.  dataset you want to evaluate:
    >> dataset = 'mnist_bgimg'; % or 'mnist_bgrand';

    2.  add liblinear library.

    3.  if your system is equipped with GPU Jacket, 
    >> usejacket = 1;
    >> gselect(1);

        if your system is not equipped with GPU Jacket,    
    >> usejacket = 0;

c.  run demo code.
    1.>> pgbm_rbm_svm this is a svm classification
    2.>> pgbm_rbm_knn this is a knn classification
    3.>> pgbm_rbm_elm this is a elm classification