Distributed deep learning on edge-devices: feasibility via adaptive compression

TL;DR

This paper explores the feasibility of distributed deep learning on edge devices by introducing adaComp, a compression algorithm that significantly reduces data transfer without sacrificing model accuracy.

Contribution

The paper presents adaComp, a novel compression method for worker updates in distributed deep learning, enabling efficient training over WAN-connected devices.

Findings

Data transfer reduced by up to 191-fold.

Model accuracy preserved despite compression.

Effective on heterogeneous and unreliable devices.

Abstract

A large portion of data mining and analytic services use modern machine learning techniques, such as deep learning. The state-of-the-art results by deep learning come at the price of an intensive use of computing resources. The leading frameworks (e.g., TensorFlow) are executed on GPUs or on high-end servers in datacenters. On the other end, there is a proliferation of personal devices with possibly free CPU cycles; this can enable services to run in users' homes, embedding machine learning operations. In this paper, we ask the following question: Is distributed deep learning computation on WAN connected devices feasible, in spite of the traffic caused by learning tasks? We show that such a setup rises some important challenges, most notably the ingress traffic that the servers hosting the up-to-date model have to sustain. In order to reduce this stress, we propose adaComp, a novel algorithm for compressing worker updates to the model on the server. Applicable to stochastic gradient descent based approaches, it combines efficient gradient selection and learning rate modulation. We then experiment and measure the impact of compression, device heterogeneity and reliability on the accuracy of learned models, with an emulator platform that embeds TensorFlow into Linux containers. We report a reduction of the total amount of data sent by workers to the server by two order of magnitude (e.g., 191-fold reduction for a convolutional network on the MNIST dataset), when compared to a standard asynchronous stochastic gradient descent, while preserving model accuracy.