Activation Map Compression through Tensor Decomposition for Deep Learning

TL;DR

This paper proposes using tensor decomposition techniques, specifically SVD and HOSVD, to compress activation maps in deep learning, significantly reducing memory use while maintaining learning performance, thus advancing edge AI capabilities.

Contribution

It introduces a novel application of tensor decomposition for activation compression in backpropagation, providing theoretical guarantees and demonstrating superior trade-offs over existing methods.

Findings

Memory savings with low-order decomposition

Preservation of essential features for learning

Pareto-superior performance on mainstream architectures

Abstract

Internet of Things and Deep Learning are synergetically and exponentially growing industrial fields with a massive call for their unification into a common framework called Edge AI. While on-device inference is a well-explored topic in recent research, backpropagation remains an open challenge due to its prohibitive computational and memory costs compared to the extreme resource constraints of embedded devices. Drawing on tensor decomposition research, we tackle the main bottleneck of backpropagation, namely the memory footprint of activation map storage. We investigate and compare the effects of activation compression using Singular Value Decomposition and its tensor variant, High-Order Singular Value Decomposition. The application of low-order decomposition results in considerable memory savings while preserving the features essential for learning, and also offers theoretical guarantees to convergence. Experimental results obtained on main-stream architectures and tasks demonstrate Pareto-superiority over other state-of-the-art solutions, in terms of the trade-off between generalization and memory footprint.