GAL: Gradient Assisted Learning for Decentralized Multi-Organization Collaborations

Authors: Enmao Diao, Jie Ding, Vahid Tarokh

NeurIPS 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental studies demonstrate that GAL can achieve performance close to centralized learning when all data, models, and objective functions are fully disclosed.
Researcher Affiliation Academia Enmao Diao Department of Electrical and Computer Engineering Duke University Durhm, NC 27705, USA enmao.diao@duke.edu Jie Ding School of Statistics University of Minnesota-Twin Cities Minneapolis, MN 55455, USA dingj@umn.edu Vahid Tarokh Department of Electrical and Computer Engineering Duke University Durhm, NC 27705, USA vahid.tarokh@duke.edu
Pseudocode Yes Algorithm 1 GAL: Gradient Assisted Learning (from the perspective of the service receiver, Alice)
Open Source Code Yes Our code is available here 1. and Did you include the code, data, and instructions needed to reproduce the main experimental results (either in the supplemental material or as a URL)? [Yes] We provide source codes in the supplementary material.
Open Datasets Yes We demonstrate the performance of autonomous local models with UCI datasets downloadable from the scikit-learn package [24], including Diabetes [25], Boston Housing [26], Blob [24], Iris [27], Wine [28], Breast Cancer [29], and QSAR [30] datasets and Our code is available here 1. We use publicly available datasets.
Dataset Splits No For all the UCI datasets, we train on 80% of the available data and test on the remaining.
Hardware Specification Yes One Nvidia 1080TI is enough for one experiment run.
Software Dependencies No We demonstrate the performance of autonomous local models with UCI datasets downloadable from the scikit-learn package [24]
Experiment Setup Yes Details of learning hyper-parameters are included in Table 9 of the Appendix. We conducted four random experiments for all datasets with different seeds, and the standard errors are shown in the brackets of all tables.