Learning without gradient descent encoded by the dynamics of a neurobiological model

TL;DR

This paper introduces a novel neurobiological model for machine learning that encodes and classifies images through dynamic signaling without gradient descent or training, achieving near state-of-the-art accuracy.

Contribution

It presents a new approach leveraging neurobiological dynamics and geometric constraints to perform image classification without traditional training methods.

Findings

Successfully encoded and classified MNIST images without training.

Achieved nearly state-of-the-art accuracy in an unsupervised manner.

Demonstrated the potential of neurobiological models for machine learning tasks.

Abstract

The success of state-of-the-art machine learning is essentially all based on different variations of gradient descent algorithms that minimize some version of a cost or loss function. A fundamental limitation, however, is the need to train these systems in either supervised or unsupervised ways by exposing them to typically large numbers of training examples. Here, we introduce a fundamentally novel conceptual approach to machine learning that takes advantage of a neurobiologically derived model of dynamic signaling, constrained by the geometric structure of a network. We show that MNIST images can be uniquely encoded and classified by the dynamics of geometric networks with nearly state-of-the-art accuracy in an unsupervised way, and without the need for any training.