Unifying back-propagation and forward-forward algorithms through model predictive control

TL;DR

This paper presents a unified framework for training neural networks by connecting back-propagation and forward-forward algorithms through model predictive control, enabling flexible trade-offs between performance and efficiency.

Contribution

It introduces an MPC-based approach that unifies BP and FF algorithms and provides a method to select optimal look-forward horizons based on specific objectives.

Findings

The framework unifies BP and FF algorithms.

Trade-offs between performance and efficiency are characterized.

Numerical results show versatility across models and tasks.

Abstract

We introduce a Model Predictive Control (MPC) framework for training deep neural networks, systematically unifying the Back-Propagation (BP) and Forward-Forward (FF) algorithms. At the same time, it gives rise to a range of intermediate training algorithms with varying look-forward horizons, leading to a performance-efficiency trade-off. We perform a precise analysis of this trade-off on a deep linear network, where the qualitative conclusions carry over to general networks. Based on our analysis, we propose a principled method to choose the optimization horizon based on given objectives and model specifications. Numerical results on various models and tasks demonstrate the versatility of our method.