PerFedMask: Personalized Federated Learning with Optimized Masking Vectors
Authors: Mehdi Setayesh, Xiaoxiao Li, Vincent W.S. Wong
ICLR 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Results based on CIFAR-10 and CIFAR-100 datasets show that the proposed Per Fed Mask algorithm provides a higher test accuracy after fine-tuning and lower average number of trainable parameters when compared with six existing state-of-the-art FL algorithms in the literature. The codes are available at https://github.com/Mehdi Set/Per Fed Mask. |
| Researcher Affiliation | Academia | Mehdi Setayesh, Xiaoxiao Li, and Vincent W.S. Wong Department of Electrical and Computer Engineering The University of British Columbia {setayeshm,xiaoxiao.li,vincentw}@ece.ubc.ca |
| Pseudocode | Yes | Algorithm 2 in Appendix B describes the Device Local Update function based on (1) and (2). (...) Algorithm 1 summarizes the training procedure of Per Fed Mask. |
| Open Source Code | Yes | The codes are available at https://github.com/Mehdi Set/Per Fed Mask. |
| Open Datasets | Yes | We conduct our experiments on CIFAR-10 and CIFAR-100 image classification tasks4. (...) 4We also use Alex Net on Domain Net dataset (Li et al., 2021) and provide the results in Appendix I to show the performance under feature non-IID configuration. |
| Dataset Splits | Yes | Each device has 450 training data samples, 50 validation data samples, and 100 test data samples. |
| Hardware Specification | No | The paper does not explicitly specify the hardware components (e.g., CPU, GPU models, memory) used for running the experiments. |
| Software Dependencies | Yes | We perform the experiments using Py Torch library (Py Torch, 2022) in Python 3.7. |
| Experiment Setup | Yes | The batch size is set to 50. (...) For all the experiments, the learning rate starts with 0.1 and is decayed by a factor of 0.1 in communication round t { 1/4T, 1/2T, 3/4T}. (...) We fix the product of the local epochs E and the maximum number of communication rounds T to 320. |