Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..
Byzantine-Resilient High-Dimensional SGD with Local Iterations on Heterogeneous Data
Authors: Deepesh Data, Suhas Diggavi
ICML 2021 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We corroborate our theoretical results with preliminary experiments for neural network training. In this section, we present preliminary numerical results on a non-convex objective. Additional implementation details can be found in Appendix F in the supplementary material. |
| Researcher Affiliation | Academia | 1University of California, Los Angeles, USA. Correspondence to: Deepesh Data <EMAIL>. |
| Pseudocode | Yes | Algorithm 1 Byzantine-Resilient SGD with Local Iterations; Algorithm 2 Robust Accumulated Gradient Estimation (RAGE) |
| Open Source Code | No | The paper mentions implementation details in Appendix F, but does not state that source code for the described methodology is being released or provide a link to it. |
| Open Datasets | Yes | We train a single layer neural network for image classification on the MNIST handwritten digit (from 0-9) dataset. |
| Dataset Splits | No | The MNIST dataset has 60,000 training images (with 6000 images of each label) and 10,000 test images (each having 28x28 = 784 pixels)... |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., GPU/CPU models, memory) used for running its experiments. |
| Software Dependencies | No | The paper does not provide specific software dependencies with version numbers. |
| Experiment Setup | Yes | All clients compute stochastic gradients on a batch-size of 128 in each iteration and communicate the local parameter vectors with the server after taking H = 7 local iterations. For all the defense mechanisms, we start with a step-size = 0.08 and decrease its learning rate by a factor of 0.96 when the difference in the corresponding test accuracies in the last 2 consecutive epochs is less than 0.001. |