Efficient On-Device Machine Learning with a Biologically-Plausible Forward-Only Algorithm

Baichuan Huang, Amir Aminifar

Conference on Machine Learning and Systems 2025 · Day 3 · Session 6: Edge and Cloud Systems

This talk introduces **BioFO (Biologically Plausible Forward-Only Algorithm)**, a novel approach to training neural networks designed to address the significant energy consumption and biological implausibility inherent in traditional backpropagation (BP) methods. Presented by Baichuan Huang from Lund University, the research highlights the urgent need for more sustainable and efficient machine learning paradigms, particularly for on-device applications. With the global average temperature rising and large language models (LLMs) like GPT-4 consuming staggering amounts of energy—48 times more than GPT-3 for training—the environmental impact of AI development is becoming a critical concern.

AI review

BioFO is a legitimate research contribution — a forward-only training algorithm that addresses real biological implausibility issues in backpropagation, with credible memory and energy efficiency results on constrained hardware. The core mechanism (local auxiliary classifiers, fixed random projections, layer-wise cross-entropy loss) is clearly explained and the Jetson Nano benchmarks ground it in something real. But this is a research talk, not an engineering talk, and the gap between 'works on CIFAR-100' and 'useful for practitioners building systems today' is never seriously addressed. The…