Reproduce “PSL-RGHA: Parallel Split Learning with Reference Gradient Half-Space Alignment for Mitigating Gradient Divergence” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Han Li, Zheng Lin, Miao Yang, Zehang Lin, Jing Yang, Zihan Fang, Qiyang Zhang, Zhaolu Kang and 4 more
PSL-RGHA addresses instability in parallel split learning under heterogeneous client data by estimating a shared reference direction from client gradients on the server and adjusting each gradient against it without aggregating client models. On non-IID CIFAR-10 with ResNet-18, it reaches 82.68% final top-1 test accuracy, compared with 79.74% for SFL; the table does not specify the Dirichlet partition parameter.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.