Reducing Communication for Split Learning by Randomized Top-k Sparsification

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zheng, Fei, Chen, Chaochao, Lyu, Lingjuan, Yao, Binhui
Format: Preprint
Published: 2023
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913206311583744
author Zheng, Fei
Chen, Chaochao
Lyu, Lingjuan
Yao, Binhui
author_facet Zheng, Fei
Chen, Chaochao
Lyu, Lingjuan
Yao, Binhui
contents Split learning is a simple solution for Vertical Federated Learning (VFL), which has drawn substantial attention in both research and application due to its simplicity and efficiency. However, communication efficiency is still a crucial issue for split learning. In this paper, we investigate multiple communication reduction methods for split learning, including cut layer size reduction, top-k sparsification, quantization, and L1 regularization. Through analysis of the cut layer size reduction and top-k sparsification, we further propose randomized top-k sparsification, to make the model generalize and converge better. This is done by selecting top-k elements with a large probability while also having a small probability to select non-top-k elements. Empirical results show that compared with other communication-reduction methods, our proposed randomized top-k sparsification achieves a better model performance under the same compression level.
format Preprint
id arxiv_https___arxiv_org_abs_2305_18469
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Reducing Communication for Split Learning by Randomized Top-k Sparsification
Zheng, Fei
Chen, Chaochao
Lyu, Lingjuan
Yao, Binhui
Machine Learning
Distributed, Parallel, and Cluster Computing
Split learning is a simple solution for Vertical Federated Learning (VFL), which has drawn substantial attention in both research and application due to its simplicity and efficiency. However, communication efficiency is still a crucial issue for split learning. In this paper, we investigate multiple communication reduction methods for split learning, including cut layer size reduction, top-k sparsification, quantization, and L1 regularization. Through analysis of the cut layer size reduction and top-k sparsification, we further propose randomized top-k sparsification, to make the model generalize and converge better. This is done by selecting top-k elements with a large probability while also having a small probability to select non-top-k elements. Empirical results show that compared with other communication-reduction methods, our proposed randomized top-k sparsification achieves a better model performance under the same compression level.
title Reducing Communication for Split Learning by Randomized Top-k Sparsification
topic Machine Learning
Distributed, Parallel, and Cluster Computing
url https://arxiv.org/abs/2305.18469