ilguyi / rbm.cpp

Restricted Boltzmann Machine

implement rbm using both Boost and Armadillo library in C++

Author

Il Gu Yi

Usage

• execute: ./rbm train.d paras.rbm.d
• argument 1: train data
• argument 2: parameters data
• DATA: MNIST (partial data)

What program can do (version 0.1) (2015. 07. 17.)

• both Binary RBM and Gaussian Bernoulli RBM
• constraint standard deviations are 1 in GBRBM case

Requirement

• I use the random number generator mt19937 from Boost library for weights and bias initialization and stochastic gradient descent.
• I implement my program using Armadillo linear algebra library in C++ for various calculation based on matrix and vector.