Unsupervised Learning with Restricted Boltzmann Machines

This is the repository of my 4th-year-seminar (a.k.a. colloquio) at Scuola Normale Superiore.

The purpose of this seminar is to illustrate the operation of Restricted Boltzmann Machines and some classical algorithms to train them. In addition, an alternative algorithm proposed by Gabriè [1] is analyzed; it is based on the mean field theory of the Ising model.

All the discussed algorithms have been implemented and tested in C++.

Documents

These are the last avaible versions of:

• Slides slides used during the presentation (version with timing)
• NoteBook personal notes and transcription from the papers that I need for prepare the talk.

References

[1] Gabrié M., Tramel E.W. and Krzakala F., 2015, December.
Training restricted Boltzmann machines via the Thouless-Anderson-Palmer free energy.
In Proceedings of the 28th International Conference on Neural Information Processing Systems-Volume 1 (pp. 640-648).