Chaos in Temperature in Diluted Mean-Field Spin-Glasses

TL;DR

This paper investigates temperature chaos in diluted mean-field spin-glass models, revealing a stronger chaos effect than in fully connected models, except for specific Bethe lattice cases, through an order parameter expansion.

Contribution

It demonstrates that diluted mean-field spin-glass models exhibit significantly stronger temperature chaos than the fully connected SK model, except on Bethe lattices with bimodal couplings.

Findings

Stronger chaos effect in diluted models compared to SK model

Exception found in Bethe lattice with bimodal couplings

Order parameter expansion used to analyze chaos

Abstract

We consider the problem of temperature chaos in mean-field spin-glass models defined on random lattices with finite connectivity. By means of an expansion in the order parameter we show that these models display a much stronger chaos effect than the fully connected Sherrington-Kirkpatrick model with the exception of the Bethe lattice with bimodal distribution of the couplings.