Towards Accurate Quantization and Pruning via Data-free Knowledge Transfer

TL;DR

This paper proposes a data-free approach for quantization and pruning of neural networks by transferring knowledge from large trained models using adversarially trained auxiliary generators, achieving competitive accuracy without access to training data.

Contribution

It introduces a novel data-free knowledge transfer method employing adversarial training and theoretical convergence analysis for quantization and pruning.

Findings

Achieves competitive accuracy with data-free methods

Demonstrates effective pruning and quantization without training data

Shows that Lottery Ticket Hypothesis aids data-free training

Abstract

When large scale training data is available, one can obtain compact and accurate networks to be deployed in resource-constrained environments effectively through quantization and pruning. However, training data are often protected due to privacy concerns and it is challenging to obtain compact networks without data. We study data-free quantization and pruning by transferring knowledge from trained large networks to compact networks. Auxiliary generators are simultaneously and adversarially trained with the targeted compact networks to generate synthetic inputs that maximize the discrepancy between the given large network and its quantized or pruned version. We show theoretically that the alternating optimization for the underlying minimax problem converges under mild conditions for pruning and quantization. Our data-free compact networks achieve competitive accuracy to networks trained and fine-tuned with training data. Our quantized and pruned networks achieve good performance while being more compact and lightweight. Further, we demonstrate that the compact structure and corresponding initialization from the Lottery Ticket Hypothesis can also help in data-free training.