Compact and Optimal Deep Learning with Recurrent Parameter Generators

TL;DR

This paper introduces a novel recurrent parameter generator that decouples model size from degrees of freedom, enabling highly compact yet accurate deep learning models through end-to-end constrained optimization.

Contribution

The paper presents a new approach using a recurrent parameter generator to optimize models with random linear constraints, achieving significant parameter reduction while maintaining high accuracy.

Findings

Achieves 96% of ResNet18 accuracy with only 18% of parameters.

Log-linear relationship between model DoF and accuracy.

Models can be pruned and quantized for further efficiency.

Abstract

Deep learning has achieved tremendous success by training increasingly large models, which are then compressed for practical deployment. We propose a drastically different approach to compact and optimal deep learning: We decouple the Degrees of freedom (DoF) and the actual number of parameters of a model, optimize a small DoF with predefined random linear constraints for a large model of arbitrary architecture, in one-stage end-to-end learning. Specifically, we create a recurrent parameter generator (RPG), which repeatedly fetches parameters from a ring and unpacks them onto a large model with random permutation and sign flipping to promote parameter decorrelation. We show that gradient descent can automatically find the best model under constraints with faster convergence. Our extensive experimentation reveals a log-linear relationship between model DoF and accuracy. Our RPG demonstrates remarkable DoF reduction and can be further pruned and quantized for additional run-time performance gain. For example, in terms of top-1 accuracy on ImageNet, RPG achieves $96\%$ of ResNet18's performance with only $18\%$ DoF (the equivalent of one convolutional layer) and $52\%$ of ResNet34's performance with only $0.25\%$ DoF! Our work shows a significant potential of constrained neural optimization in compact and optimal deep learning.