Performative Federated Learning: A Solution to Model-Dependent and Heterogeneous Distribution Shifts

Authors: Kun Jin, Tongxin Yin, Zhongzhu Chen, Zeyu Sun, Xueru Zhang, Yang Liu, Mingyan Liu

AAAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Numerical results validate our analysis and provide valuable insights into real-world applications. We show empirically that P-Fed Avg also has good performance on FMNIST (Deng 2012) and Cifar-10.
Researcher Affiliation Academia 1 University of Michigan 2 The Ohio State University 3 University of California, Santa Cruz
Pseudocode No The paper describes the P-Fed Avg algorithm in Section 2.4, but it does not include a formally structured 'Algorithm' or 'Pseudocode' block.
Open Source Code Yes Our code is publicly accessible1. 1https://github.com/tsy19/Performative Fed Avg
Open Datasets Yes Demonstrating the efficacy of P-Fed Avg on the Kaggle dataset2 as per Perdomo et al. (2020). We show empirically that P-Fed Avg also has good performance on FMNIST (Deng 2012) and Cifar-10.
Dataset Splits No The paper mentions 'a 10% subset of the training data' for each client in the Credit Score Strategic Classification, but does not provide specific train/validation/test split percentages or absolute counts for any of the datasets used (Kaggle, FMNIST, Cifar-10).
Hardware Specification No The paper does not provide specific hardware details such as GPU/CPU models, processor types, or memory amounts used for running its experiments.
Software Dependencies No The paper does not provide specific ancillary software details, such as library or solver names with version numbers, needed to replicate the experiments.
Experiment Setup Yes For the Credit Score Strategic Classification experiment, the paper states using 'partial participation with K = 5', '5 gradient descent steps per round', and 'a minibatch of size 4'. For Performative Image Classification it states 'β = 0.5'.