Controlling energy landscapes with correlations between minima

TL;DR

This paper demonstrates that the dynamic properties of neural network energy landscapes, including basin sizes and state densities, can be manipulated through the correlations and hierarchical structures of stored memory patterns.

Contribution

It introduces a method to control energy landscape dynamics by adjusting correlations between memory patterns in neural networks.

Findings

Dynamic properties depend on pattern correlations

Hierarchical structures alter basin sizes and state densities

Energy landscape features can be tuned by pattern correlations

Abstract

Neural network models have been used to construct energy landscapes for modeling biological phenomena, in which the minima of the landscape correspond to memory patterns stored by the network. Here, we show that dynamic properties of those landscapes, such as the sizes of the basins of attraction and the density of stable and metastable states, depend strongly on the correlations between the memory patterns and can be altered by introducing hierarchical structures. Our findings suggest dynamic features of energy landscapes can be controlled by choosing the correlations between patterns