Finding Local Minima Efficiently in Decentralized Optimization

Authors: Wenhan Xian, Heng Huang

NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In this section we will demonstrate our experimental results to validate the performance of our method. We conduct two tasks in our experiment, a matrix sensing task with synthetic data and a matrix factorization task with a real-world dataset to validate the performance of our method.
Researcher Affiliation Academia Wenhan Xian Computer Science University of Maryland College Park wxian1@umd.edu Heng Huang Computer Science University of Maryland College Park heng@umd.edu
Pseudocode Yes Algorithm 1 Perturbed Decentralized STORM Algorithm (PEDESTAL)
Open Source Code Yes The source code is available in https://github.com/WH-XIAN/PEDESTAL.
Open Datasets Yes We conduct two tasks in our experiment, a matrix sensing task with synthetic data and a matrix factorization task with a real-world dataset... solve this problem on the Movie Lens-100k dataset (Harper and Konstan [2015]).
Dataset Splits No The paper describes how data is generated and assigned to worker nodes but does not specify explicit training, validation, or test dataset splits with percentages or sample counts.
Hardware Specification No The paper does not provide specific hardware details such as GPU/CPU models, memory, or cloud instance types used for running the experiments.
Software Dependencies No The paper does not list specific software dependencies with version numbers (e.g., libraries, frameworks, or programming language versions) used for implementing or running the experiments.
Experiment Setup Yes In this experiment, the learning rate is chosen from {0.01, 0.001, 0.0001}. The batchsize is set to 10. For PEDESTAL and GTHSGD, parameter β is set to 0.01. For D-GET and D-SPIDER-SFO, the period q is 100. For PEDESTAL, threshold Cv is set to 0.0001. Perturbation radius r is set to 0.001. The threshold of moving distance Cd is set to 0.01.