Sparse Hopfield network reconstruction with $\ell_{1}$ regularization

TL;DR

This paper introduces an efficient method for reconstructing sparse Hopfield networks using $\, ext{l}_1$ regularization within the Bethe approximation, improving inference accuracy with minimal computational effort.

Contribution

It develops a novel $\, ext{l}_1$ regularization approach integrated into the Bethe approximation for sparse network inference, applicable to various diluted mean field models.

Findings

Regularization parameter scales as $M^{- u}$ with $ u$ depending on the performance measure.

Method reduces inference error compared to Bethe approximation without regularization.

Demonstrated effectiveness on sparse Hopfield models with low computational cost.

Abstract

We propose an efficient strategy to infer sparse Hopfield network based on magnetizations and pairwise correlations measured through Glauber samplings. This strategy incorporates the $\ell_{1}$ regularization into the Bethe approximation by a quadratic approximation to the log-likelihood, and is able to further reduce the inference error of the Bethe approximation without the regularization. The optimal regularization parameter is observed to be of the order of $M^{-\nu}$ where $M$ is the number of independent samples. The value of the scaling exponent depends on the performance measure. $\nu\simeq0.5001$ for root mean squared error measure while $\nu\simeq0.2743$ for misclassification rate measure. The efficiency of this strategy is demonstrated for the sparse Hopfield model, but the method is generally applicable to other diluted mean field models. In particular, it is simple in implementation without heavy computational cost.