Adaptive Multi-Scale Goodness Aggregation for Forward-Forward Learning

TL;DR

This paper introduces AMSGA, an extension of the Forward-Forward algorithm, enhancing stability, robustness, and generalization in local-learning neural networks through multi-scale aggregation and adaptive training strategies.

Contribution

It presents AMSGA, a novel method that improves the Forward-Forward algorithm by incorporating multi-scale goodness aggregation and adaptive training techniques.

Findings

Achieves up to +1.45% accuracy improvement on MNIST.

Achieves up to +1.50% accuracy improvement on Fashion-MNIST.

Demonstrates improved stability and robustness in local-learning neural networks.

Abstract

We propose Adaptive Multi-Scale Goodness Aggregation (AMSGA), a novel extension of the Forward-Forward (FF) algorithm designed to improve stability, robustness, and generalization in local-learning neural networks. AMSGA addresses several limitations of the original FF framework by introducing multi-scale goodness aggregation across local, intermediate, and global representations; adaptive curriculum-guided hard negative mining; layer-dependent adaptive thresholds; and a warm-up cosine annealing learning-rate schedule for improved optimization stability. Together, these modifications strengthen the FF paradigm while preserving its biologically plausible and memory-efficient properties. Experiments on MNIST and Fashion-MNIST demonstrate consistent performance improvements over the baseline FF algorithm, achieving up to +1.45% improvement on MNIST and +1.50% improvement on Fashion-MNIST without significant computational overhead. Our results suggest that local learning methods can become substantially more competitive when goodness estimation and training dynamics are carefully designed.