Towards More Efficient Stochastic Decentralized Learning: Faster Convergence and Sparse Communication

Authors: Zebang Shen, Aryan Mokhtari, Tengfei Zhou, Peilin Zhao, Hui Qian

ICML 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments on convex minimization and AUC-maximization validate the efficiency of our method. and In this section, we evaluate the empirical performance of DSBA and compare it with several state-of-the-art methods including: DSA, EXTRA, SSDA, and DLM.
Researcher Affiliation Collaboration 1Zhejiang University 2Tencent AI Lab 3Massachusetts Institute of Technology 4South China University of Technology.
Pseudocode Yes Algorithm 1 DSBA for node n and Algorithm 2 Computation on node 0 at iteration t
Open Source Code No The paper does not provide concrete access to source code or a statement about its public availability.
Open Datasets Yes As to dataset, we use News20-binary, RCV1, and Sector from LIBSVM dataset
Dataset Splits No The paper states it "randomly split the them into N partitions with equal sizes" but does not provide specific train/validation/test split percentages, sample counts, or references to predefined splits.
Hardware Specification No The paper does not provide specific hardware details (e.g., CPU, GPU models, or memory) used for running the experiments.
Software Dependencies No The paper does not provide specific details regarding software dependencies with version numbers (e.g., programming languages, libraries, or frameworks).
Experiment Setup Yes We tune the step size of all algorithms and select the ones that give the best performance. and The ℓ2-regularization parameter λ is set to 1/(10Q) in all cases.