Recognition: unknown
Metastability in the generalized Hopfield model with finitely many patterns
read the original abstract
This paper continues the study of metastable behaviour in disordered mean field models initiated in [2], [3]. We consider the generalized Hopfield model with finitely many independent patterns $\xi_1,...,\xi_p$ where the patterns have i.i.d. components and follow discrete distributions on $[-1,1]$. We show that metastable behaviour occurs and provide sharp asymptotics on metastable exit times and the corresponding capacities. We apply the potential theoretic approach developed by Bovier et al. in the space of appropriate order parameters and use an analysis of the discrete Laplacian to obtain lower bounds on capacities. Moreover, we include the possibility of multiple saddle points with the same value of the rate function and the case that the energy surface is degenerate around critical points.
This paper has not been read by Pith yet.
Forward citations
Cited by 1 Pith paper
-
Hopfield Networks is All You Need
Modern Hopfield networks store exponentially many patterns, retrieve them in one update, and have an update rule equivalent to transformer attention, enabling new Hopfield layers that improve results on multiple insta...
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.