MPF is a technique for parameter estimation in un-normalized probabilistic models. It is described in the paper:
J Sohl-Dickstein, P Battaglino, MR DeWeese
Minimum probability flow learning
International Conference on Machine Learning (2011)
http://arxiv.org/abs/0906.4779
This repository contains Matlab code implementing MPF for the Ising model and the RBM. The directory structure is as follows:
- MPF_ising/ - parameter estimation in the Ising model
- MPF_RBM_compare_log_likelihood/ - parameter estimation in Restricted Boltzmann Machines. This directory also includes code comparing the log likelihood of small RBMs trained via pseudolikelihood and Contrastive Divergence to ones trained via MPF.
If you're interesting in using MPF to build an Ising model of neural spike data, you should also check out Liberty Hamilton's repository at https://github.com/libertyh/ising-model.