Generalisation to unseen topologies: Towards control of biological neural network activity

TL;DR

This paper introduces a procedurally generated environment for neuronal networks and evaluates a transformer-based deep reinforcement learning agent's ability to generalize control across unseen network topologies, advancing adaptive neural control.

Contribution

It presents a novel environment for testing control generalization in biological neural networks and adapts a transformer-based architecture for this purpose.

Findings

The agent successfully generalizes control to unseen network topologies.

The environment enables testing of control strategies in diverse neuronal structures.

Transformer-based models show promise in adaptive neural network control.

Abstract

Novel imaging and neurostimulation techniques open doors for advancements in closed-loop control of activity in biological neural networks. This would allow for applications in the investigation of activity propagation, and for diagnosis and treatment of pathological behaviour. Due to the partially observable characteristics of activity propagation, through networks in which edges can not be observed, and the dynamic nature of neuronal systems, there is a need for adaptive, generalisable control. In this paper, we introduce an environment that procedurally generates neuronal networks with different topologies to investigate this generalisation problem. Additionally, an existing transformer-based architecture is adjusted to evaluate the generalisation performance of a deep RL agent in the presented partially observable environment. The agent demonstrates the capability to generalise control from a limited number of training networks to unseen test networks.