Active Inference Demonstrated with Artificial Spin Ice

TL;DR

This paper demonstrates that a bilayer artificial spin ice system can perform active inference driven by thermal fluctuations, aligning with neurological free energy principles, and reproduces biological neural process models.

Contribution

It introduces a novel bilayer magnetic nanostructure that models active inference, bridging neuroscience theories and nanomagnetic systems.

Findings

Thermal fluctuations drive the spin system under external constraints.

The bilayer structure reproduces active inference behaviors.

The model aligns with neurological free energy principles.

Abstract

A numerical model of interacting nanomagnetic elements is used to demonstrate active inference with a three dimensional Artificial Spin Ice structure. It is shown that thermal fluctuations can drive this magnetic spin system to evolve under dynamic constraints imposed through interactions with an external environment as predicted by the neurological free energy principle and active inference. The structure is defined by two layers of magnetic nanoelements where one layer is a square Artificial Spin Ice geometry. The other magnetic layer functions as a sensory filter that mediates interaction between the external environment and the hidden Artificial Spin Ice layer. Spin dynamics displayed by the bilayer structure are shown to be well described using a continuous form of a neurological free energy principle that has been previously proposed as a high level description of certain biological neural processes. Numerical simulations demonstrate that this proposed bilayer geometry is able to reproduce theoretical results derived previously for examples of active inference in neurological contexts.