Example Paper 1: A Novel Approach to Synaptic Plasticity

Note: This is a placeholder entry to demonstrate the layout and structure of the ONR Approved Research Registry.

Abstract

The backpropagation of error algorithm is arguably the most important algorithm in artificial intelligence, but has been deemed biologically implausible. In this placeholder study, we introduce a novel, biologically plausible learning rule inspired by the dynamics of astrocytic networks. Our model, termed Astro-Modulated Hebbian Learning (AMHL), demonstrates competitive performance with backpropagation on standard image classification benchmarks while requiring only local information for weight updates. We validate our findings through simulations on synthetic and real-world datasets, suggesting a viable path toward more brain-like on-chip learning.

Resource Overview

This resource consists of a pre-print manuscript and an accompanying open-source Python codebase. The codebase, implemented in PyTorch and snnTorch, allows researchers to replicate all experiments presented in the paper. It includes scripts for data preprocessing, model training, and result visualization. We hope this work serves as a foundation for further exploration into glial-neural interactions in computational models.