Supervised Compression for Split Computing

Summary

Split computing has been attracting attentions from the research community as an intermediate option between local (mobile) computing and full computation offloading. This approach is effective especially for resource-constrained edge computing systems e.g., computationally weak mobile devices, limited wireless communication capacity. We proposed introducing bottleneck (splitting point) to deep learning models at their early stages so that we can minimize local computing and data communication costs while maximizing model accuracy, which we formulated later as supervised compression for split computing (SC2).

We define supervised compression as learning compressed representations for supervised downstream tasks. It is very challenging to introduce bottlenecks (splitting points) to deep learning models at their early layers while preserving model accuracy. SC2 is a key to achieve efficient and accuracy machine learning applications for resource-constrained edge computing systems.

ML OSS

To facilitate research on supervised compression for split computing (SC2), I published an installable Python package named sc2bench (i.e., "pip3 install sc2bench") and a code repository to reproduce the experimental results reported in our SC2 paper.

Related Publications

A Multi-task Supervised Compression Model for Split Computing

WACV 2025

Yoshitomo Matsubara, Matteo Mendula, Marco Levorato

This work was done prior to joining Spiffy AI.

Preprint Code

SC2 Benchmark: Supervised Compression for Split Computing

Transactions on Machine Learning Research (TMLR)

Yoshitomo Matsubara, Ruihan Yang, Marco Levorato, Stephan Mandt

This work was done prior to joining Amazon.

OpenReview Video Preprint Code PyPI

SplitBeam: Effective and Efficient Beamforming in Wi-Fi Networks Through Split Computing

ICDCS 2023

Niloofar Bahadori, Yoshitomo Matsubara, Marco Levorato, Francesco Restuccia

This work was done prior to joining Amazon.

Paper Preprint Code Dataset

Split Computing and Early Exiting for Deep Learning Applications: Survey and Research Challenges

ACM Computing Surveys (CSUR)

Yoshitomo Matsubara, Marco Levorato, Francesco Restuccia

Paper Preprint

BottleFit: Learning Compressed Representations in Deep Neural Networks for Effective and Efficient Split Computing

IEEE WoWMoM 2022

Yoshitomo Matsubara, Davide Callegaro, Sameer Singh, Marco Levorato, Francesco Restuccia

Paper Preprint Code

Supervised Compression for Resource-Constrained Edge Computing Systems

WACV 2022

Yoshitomo Matsubara, Ruihan Yang, Marco Levorato,

Paper Preprint Code

Neural Compression and Filtering for Edge-assisted Real-time Object Detection in Challenged Networks

ICPR 2020

Yoshitomo Matsubara, Marco Levorato

Paper Supp Preprint Code

Optimal Task Allocation for Time-Varying Edge Computing Systems with Split DNNs

IEEE GLOBECOM 2020

Davide Callegaro, Yoshitomo Matsubara, Marco Levorato

Paper

Head Network Distillation: Splitting Distilled Deep Neural Networks for Resource-constrained Edge Computing Systems

IEEE Access

Yoshitomo Matsubara, Davide Callegaro, Sabur Baidya, Marco Levorato, Sameer Singh

Paper Code

Split Computing for Complex Object Detectors: Challenges and Preliminary Results

MobiCom 2020 Workshop on Embedded and Mobile Deep Learning (EMDL)

Yoshitomo Matsubara, Marco Levorato

Paper Preprint Code

Distilled Split Deep Neural Networks for Edge-Assisted Real-Time Systems

MobiCom 2019 Workshop on Hot Topics in Video Analytics and Intelligent Edges (HotEdgeVideo)

Yoshitomo Matsubara, Sabur Baidya, Davide Callegaro, Marco Levorato, Sameer Singh

Paper Code