Skip to content

Feature Map Compression for Split Learning in Mobile Wireless Environment: Theory and Practice

Oct 2026 · IEEE Transactions on Mobile Computing · Vol 25, pp. 17520-17534 · 0 citations · 50 references

Abstract

Split Learning (SL) enables collaborative model training by distributing computation between resource-constrained mobile edge devices and a central server. In this distributed framework, mobile edge devices compute the early layers of the model and transmit feature maps to the server. However, due to practical mobile wireless bandwidth limitations, these feature maps should be compressed before transmission. In this paper, we propose a lightweight Mask-Encoded Sparsification (MS) compressor that enhances Top-<inline-formula><tex-math notation="LaTeX">$k$</tex-math><alternatives><mml:math><mml:mi>k</mml:mi></mml:math><inline-graphic xlink:href="qu-ieq1-3696694.gif"/></alternatives></inline-formula> sparsification with a narrow-bit mask, significantly reducing compression error while improving computational and communication efficiency. Our theoretical analyses reveal that feature map compression generally induces estimator bias in the gradients. Furthermore, we establish the relationships between compression error and convergence, as well as between the choice of cutlayer and output error. The first relationship demonstrates that reducing compression error facilitates convergence, confirming the advantage of MS, whereas the second provides theoretical guidance for determining the cutlayer, indicating that earlier layers are more sensitive to compression errors. Extensive experiments on vision and language models demonstrate that our method outperforms existing compression techniques in terms of compression error and model performance, achieving up to a 21.38× speedup in communication efficiency under wireless environments.

View source

Similar papers

2026

Computation-Aware Adaptive and Scalable Deep Joint Source-Channel Coding for Heterogeneous Broadcast

In recent years, Deep Joint Source-Channel Coding (DeepJSCC) has shown significant advantages in wireless image transmission. However, in practical broadcast scenarios, receivers often exhibit substantial heterogeneity in computational capability, bandwidth, and reconstruction requirements, which poses major challenges...

Feng Wang, Xuechen Chen, Jia-Qi Liu et al. · 0 citations
Open access 2026

Phase-Aware Quantum–Classical Feedback Compression in RIS: Trainability and Performance Limits

This paper considers a multi-antenna base station (BS) communicating with a single-antenna user through a reconfigurable intelligent surface (RIS). In large RIS deployments, feeding back the high-dimensional phase vector <inline-formula><tex-math notation="LaTeX">$\boldsymbol{\theta }\in [0,2\pi)^{N}$</tex-math></inlin...

Sudhanshu Arya · 0 citations
Conference Aug 2026

ASC-Net: Deep Learning for Downlink Control Information Compression

With the rapid growth of connected devices in the sixth generation (6G) era, the overhead of control signaling faces severe pressure. Consequently, downlink control information (DCI) compression becomes an effective approach to alleviate control channel congestion while maintaining scheduling flexibility. To achieve th...

Hanqing Xu, Long Zhao, Hong-Rui Shen et al. · 0 citations
Preprint Aug 2026

Single-Model Adaptive Wireless Image Transmission via Feature Sparsity Regularization

Learned joint source-channel coding (JSCC) enables robust wireless image transmission by jointly optimizing the transmitter and receiver over differentiable channel models. For bandwidth-limited and time-varying visual links, a single model should support user-adjustable transmission rate and adapt to changing wireless...

Xianghao Cui, Li Lan, Qi He et al. · 0 citations
2026

SpatialMix-JSCC: Spatial Mixing State Space Duality for Adaptive Wireless Image Transmission

SSMs offer linear-complexity modeling for JSCC-based wireless image transmission, but 1-D token serialization weakens 2-D spatial priors and structural fidelity, especially at low CBRs. We propose SpatialMix-JSCC, an SSM-based JSCC framework that injects 2-D spatial awareness into SSM parameterization via SpatialMix-SS...

Bo-Rui Zhang, Qipeng Chen, Wen-Tao Wang et al. · 0 citations

Related blog posts

MIT News · Artificial Intelligence Nov 17, 2026

Learning the language of machine learning

By studying the computational foundations of language learning, Associate Professor Jacob Andreas builds intelligent systems that can communicate effectively with humans.

Microsoft Research Blog Oct 29, 2026

Microsoft Research Focus – Edition #97

Host Sowbhagya Suresh shares some of the latest news and cutting-edge research from across Microsoft—from understanding the challenges facing coding agents to uncovering how people use AI at work and improving extended-range weather forecasting. The post Microsoft Research Focus – Edition #97 appeared first on Microsoft Research.

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.