Practical Network Acceleration with Tiny Sets: Hypothesis, Theory, and Algorithm

TL;DR

This paper presents a theoretical framework and an algorithm for accelerating neural networks with tiny training sets, demonstrating significant improvements over previous methods in terms of latency reduction and robustness.

Contribution

It introduces the finetune convexity hypothesis, a new theory for few-shot compression, and the PRACTISE algorithm for effective network acceleration with small data.

Findings

PRACTISE achieves 7% higher accuracy in latency reduction on ImageNet-1k.

Feature mimicking results in lower parameter variance and easier optimization.

Dropping blocks is theoretically superior for few-shot network compression.

Abstract

Due to data privacy issues, accelerating networks with tiny training sets has become a critical need in practice. Previous methods achieved promising results empirically by filter-level pruning. In this paper, we both study this problem theoretically and propose an effective algorithm aligning well with our theoretical results. First, we propose the finetune convexity hypothesis to explain why recent few-shot compression algorithms do not suffer from overfitting problems. Based on it, a theory is further established to explain these methods for the first time. Compared to naively finetuning a pruned network, feature mimicking is proved to achieve a lower variance of parameters and hence enjoys easier optimization. With our theoretical conclusions, we claim dropping blocks is a fundamentally superior few-shot compression scheme in terms of more convex optimization and a higher acceleration ratio. To choose which blocks to drop, we propose a new metric, recoverability, to effectively measure the difficulty of recovering the compressed network. Finally, we propose an algorithm named PRACTISE to accelerate networks using only tiny training sets. PRACTISE outperforms previous methods by a significant margin. For 22% latency reduction, it surpasses previous methods by on average 7 percentage points on ImageNet-1k. It also works well under data-free or out-of-domain data settings. Our code is at https://github.com/DoctorKey/Practise