The Hopfield model of neural networks

Statistical Physics and Complexity Group meeting

The Hopfield model of neural networks

  • Event time: 11:30am until 12:30pm
  • Event date: 6th November 2019
  • Speaker: Francesco Cagnetta (Formerly School of Physics & Astronomy, University of Edinburgh)
  • Location: Room 2511,

Event details

I will use the Hopfield model as a common thread to review some aspects of the statistical mechanics of neural networks. Starting from the definition of the model and connection with spin glasses, I will discuss it's representation as a restricted Boltzmann machine and how, within the latter representation, one can witness the emergence of the layered structure typical of deep learning methods.

The talk will serve as a presentation of the reading group---another weekly meeting focused on an aspect of theoretical statistical mechanics.