mlsanityData to 14 Oct 2026

Concurrent Split Learning Through Stable Client Clustering

arXiv 2609.29395 ↗First posted 24 Sep 2026cs.DC cs.LGcited by 0

Not on the map. Why?

Abstract (arXiv)

Training with a fixed global batch limits how many distributed clients can provide examples in any one step. We examine a way to use additional server workers without increasing the batch processed by an individual workload. Global Clustered Parallel Split Learning (GCPSL) assigns clients to fixed clusters, executes a Parallel Split Learning with Global Sampling (GPSL) workload for each cluster concurrently, and periodically fuses the client and server model segments. In simulations with 256 logical clients, dividing the population across more workloads improves direct data participation, while smaller clusters can incur an accuracy cost. A four-H100 implementation of label-aware GCPSL reaches 85% CIFAR-10 validation accuracy in 6.13 ± 0.15 minutes over three matched runs, versus 19.09 ± 0.45 minutes when the same workloads are serialized. Within the four-GPU allocation, size-balanced and random fixed affiliations reach the target in similar mean times (5.70 and 5.66 minutes); size balancing increases direct participation by 3.25 percentage points. These measurements characterize a trade-off among execution concurrency, assignment information, participation, and accuracy for stable-client split learning.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 25 of its 36 references that are in the map. First posted on arXiv 24 Sep 2026.

Builds on

25 of 36 references.

References · 25

  1. 2015Deep Residual Learning for Image Recognitioncited by 52,096
  2. 2016Communication-Efficient Learning of Deep Networks from Decentralized Datacited by 7,669
  3. 2018Federated Optimization in Heterogeneous Networkscited by 2,987
  4. 2018Group Normalizationcited by 1,524
  5. 2017Accurate, Large Minibatch SGD: Training ImageNet in 1 Hourcited by 1,492
  6. 2016On Large-Batch Training for Deep Learning: Generalization Gap and Sharp Minimacited by 1,352
  7. 2019Measuring the Effects of Non-Identical Data Distribution for Federated Visual Classificationcited by 851
  8. 2018LEAF: A Benchmark for Federated Settingscited by 792
  9. 2016Federated Optimization: Distributed Machine Learning for On-Device Intelligencecited by 564
  10. 2019Clustered Federated Learning: Model-Agnostic Distributed Multi-Task Optimization under Privacy Constraintscited by 503
  11. 2017Deep Gradient Compression: Reducing the Communication Bandwidth for Distributed Trainingcited by 478
  12. 2018Split learning for health: Distributed deep learning without sharing raw patient datacited by 395
13 more
  1. 2020SplitFed: When Federated Learning Meets Split Learningcited by 310
  2. 2018Distributed learning of deep neural network over multiple agentscited by 279
  3. 2019Edge Intelligence: Paving the Last Mile of Artificial Intelligence with Edge Computingcited by 228
  4. 2018Measuring the Effects of Data Parallelism on Neural Network Trainingcited by 193
  5. 2021FedScale: Benchmarking Model and System Performance of Federated Learning at Scalecited by 136
  6. 2020Unleashing the Tiger: Inference Attacks on Split Learningcited by 105
  7. 2020TiFL: A Tier-based Federated Learning Systemcited by 96
  8. 2020Communication-Efficient Edge AI: Algorithms and Systemscited by 96
  9. 2023Efficient Parallel Split Learning over Resource-constrained Wireless Edge Networkscited by 95
  10. 2022Split Learning over Wireless Networks: Parallel Design and Resource Managementcited by 75
  11. 2023MCR-DL: Mix-and-Match Communication Runtime for Deep Learningcited by 4
  12. 2024QSync: Quantization-Minimized Synchronous Distributed Training Across Hybrid Devicescited by 3
  13. 2024Parallel Split Learning with Global Samplingcited by 3

Built on by

Not cited yet.