Kumar Behera, Agnish - University of Chicago

Correspondence between Hebbian Unlearning and Non-equilibrium steady states

In a previous work (arXiv:2203.03024), we demonstrate that an associative memory system with active dynamics can improve the system’s information storage and retrieval properties. Here, we propose an equivalence between the Hebbian Unlearning algorithm (also known as “dreaming”) and modifying dynamics by driving the spins with exponentially correlated noise. We demonstrate that this is similar to how neurons integrate information and fire. Our method shows how neurons can store new memories through different dynamics, without altering existing connections.