| Documentation | Build Status |
|---|---|
 |
  |
This package implements the Focusing Belief Propagation algorithm for committee machines with binary weights described in the paper Unreasonable Effectiveness of Learning Neural Networks: From Accessible States and Robust Ensembles to Basic Algorithmic Schemes by Carlo Baldassi, Christian Borgs, Jennifer Chayes, Alessandro Ingrosso, Carlo Lucibello, Luca Saglietti and Riccardo Zecchina, Proc. Natl. Acad. Sci. U.S.A. 113: E7655-E7662 (2016), doi:10.1073/pnas.1608103113.
The code is written in Julia. It was last tested with Julia version 1.4.
To install the module, switch to pkg mode with the ] key and use this command:
pkg> add https://github.com/carlobaldassi/BinaryCommitteeMachineFBP.jl
Dependencies will be installed automatically.