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..
Redefining Contributions: Shapley-Driven Federated Learning
Authors: Nurbek Tastan, Samar Fares, Toluwani Aremu, Samuel Horváth, Karthik Nandakumar
IJCAI 2024 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experiments on CIFAR-10, Chest X-Ray, and Fed-ISIC2019 datasets demonstrate the effectiveness of our approach in improving utility, efficiency, and fairness in FL systems. |
| Researcher Affiliation | Academia | Nurbek Tastan , Samar Fares , Toluwani Aremu , Samuel Horvath , Karthik Nandakumar Mohamed bin Zayed University of Artificial Intelligence (MBZUAI), Abu Dhabi, UAE EMAIL |
| Pseudocode | Yes | Algorithm 1 Shap Fed algorithm |
| Open Source Code | Yes | The code can be found at https://github.com/tnurbek/shapfed. |
| Open Datasets | Yes | CIFAR-10 [Krizhevsky et al., 2009]: This dataset comprises 60,000 RGB images, each with dimensions of 32 × 32 pixels, spanning 10 different classes. It is divided into a training set of 50,000 images and a testing set of 10,000 images. Chest X-Ray [Rahman et al., 2020]: The Tuberculosis (TB) Chest X-ray Database is a comprehensive collection of chest X-ray images containing 700 publicly accessible TB-positive images and 3500 normal images. Fed-ISIC2019 [Ogier du Terrail et al., 2022]: This dataset is an amalgamation of the ISIC 2019 challenge dataset and the HAM1000 database, presenting a total of 23,247 dermatological images of skin lesions (8 classes). |
| Dataset Splits | No | While CIFAR-10's train/test split is mentioned, explicit and detailed validation splits or methodologies for creating them are not provided for all datasets (e.g., Chest X-Ray, Fed-ISIC2019) in the main text for full reproducibility. The mention of an "auxiliary validation set" is conceptual, not a description of its use in their experimental setup details. |
| Hardware Specification | No | The paper mentions models like Res Net-34 and Efficient Net B0, but it does not provide specific hardware details such as GPU models, CPU models, or memory specifications used for running the experiments. |
| Software Dependencies | No | The paper mentions using SGD optimizer and specific model architectures (Res Net-34, Efficient Net B0) but does not provide specific version numbers for any software libraries, frameworks (e.g., PyTorch, TensorFlow), or programming languages used. |
| Experiment Setup | Yes | CIFAR-10. We leverage the Res Net-34 architecture trained using the SGD optimizer with a fixed learning rate of 0.01. For FL, we use 50 communication rounds. Chest X-Ray. ... We use SGD optimizer with a learning rate of 0.01, momentum of 0.9, and weight decay of 5 × 10−4. The models are trained for 50 rounds. Fed-ISIC2019. We employ the Efficient Net B0 model and use the same training settings as in CIFAR-10 with 200 communication rounds. |