Skip to content
A new training strategy for the infinite RBMs.
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
AIS_RBM
AIS_iRBM
core_modules
evaluation
models
BinaryDataMNIST.mat
README.md
best_Dis_iRBM.mat
best_Dis_iRBM_0521002_NR_145.mat
caltech101_silhouettes_28_split1.mat
demo_Dis_iRBM_classification_CalTech101.m
demo_Dis_iRBM_classification_MNIST.m
demo_iRBM_density_estimation_CalTech101.m
demo_iRBM_density_estimation_MNIST.m
initiate.m

README.md

RP-iRBM

A new training strategy to the infinite RBMs.

The key idea of the proposed training strategy is randomly regrouping the hidden units before each gradient descent step.

Potentially, a mixing of infinite many iRBMs with different permutations of the hidden units can be achieved by this learning method.

The original iRBM is also modified to be capable of carrying out discriminative training(Discriminative iRBM).

For more details, please see our paper at Arxiv. https://arxiv.org/abs/1709.03239

Dependencies

Matlab 2016a or higher.

Usage

Run initiate.m to set up the environment.

Run demo_iRBM_density_estimation_MNIST.m to train an iRBM using RP on the MNIST.

Run demo_Dis_iRBM_classification_MNIST.m to train a Dis-iRBM using RP on the MNIST.

You can’t perform that action at this time.