Unified Stochastic Framework for Neural Network Quantization and Pruning

TL;DR

This paper presents a unified stochastic framework that simultaneously addresses neural network quantization and pruning, providing theoretical guarantees and extending applicability to low-bit regimes.

Contribution

It introduces a novel stochastic path-following framework that unifies quantization and pruning with rigorous error bounds and robustness, including 1-bit regimes.

Findings

Provides theoretical error bounds for quantization and pruning.

Extends SPFQ to include pruning and low-bit quantization.

Achieves robust error correction in neural network compression.

Abstract

Quantization and pruning are two essential techniques for compressing neural networks, yet they are often treated independently, with limited theoretical analysis connecting them. This paper introduces a unified framework for post-training quantization and pruning using stochastic path-following algorithms. Our approach builds on the Stochastic Path Following Quantization (SPFQ) method, extending its applicability to pruning and low-bit quantization, including challenging 1-bit regimes. By incorporating a scaling parameter and generalizing the stochastic operator, the proposed method achieves robust error correction and yields rigorous theoretical error bounds for both quantization and pruning as well as their combination.