Iteratively Training Look-Up Tables for Network Quantization

TL;DR

This paper introduces LUT-Q, a flexible framework for network reduction that learns value dictionaries and assignment matrices iteratively, enabling various quantization and pruning strategies adaptable to different hardware constraints.

Contribution

The paper presents a novel LUT-Q method that unifies multiple network reduction techniques through a single iterative learning framework.

Findings

LUT-Q effectively performs non-uniform quantization and pruning.

The method adapts to different hardware constraints.

LUT-Q achieves competitive network compression results.

Abstract

Operating deep neural networks (DNNs) on devices with limited resources requires the reduction of their memory as well as computational footprint. Popular reduction methods are network quantization or pruning, which either reduce the word length of the network parameters or remove weights from the network if they are not needed. In this article we discuss a general framework for network reduction which we call `Look-Up Table Quantization` (LUT-Q). For each layer, we learn a value dictionary and an assignment matrix to represent the network weights. We propose a special solver which combines gradient descent and a one-step k-means update to learn both the value dictionaries and assignment matrices iteratively. This method is very flexible: by constraining the value dictionary, many different reduction problems such as non-uniform network quantization, training of multiplierless networks, network pruning or simultaneous quantization and pruning can be implemented without changing the solver. This flexibility of the LUT-Q method allows us to use the same method to train networks for different hardware capabilities.