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 [1].
PerFedRLNAS: One-for-All Personalized Federated Neural Architecture Search
Authors: Dixi Yao, Baochun Li
AAAI 2024 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In our experiments, we empirically show that our framework shows much better performance with respect to personalized accuracy and overall time compared to state-of-the-art methods. We empirically compare Per Fed RLNAS with the state-of-the-art personalized federated learning methods and previous federated neural architecture works to see how well our method solves the heterogeneous problems. Dataset, tasks, and models. We study on image classification tasks with CIFAR10 and CIFAR100 (Krizhevsky, Hinton et al. 2009). |
| Researcher Affiliation | Academia | University of Toronto EMAIL, EMAIL |
| Pseudocode | Yes | Algorithm 1: Per Fed RLNAS |
| Open Source Code | Yes | Our source code is released in https://github.com/TL-System/plato/tree/main/ examples/model search/pfedrlnas. |
| Open Datasets | Yes | We study on image classification tasks with CIFAR10 and CIFAR100 (Krizhevsky, Hinton et al. 2009). |
| Dataset Splits | No | The paper mentions 'training samples and test samples are equally partitioned over all the clients' but does not explicitly provide details about a validation set or specific split percentages for train/validation/test splits. |
| Hardware Specification | No | The paper mentions 'physical (GPU) memory' but does not provide specific details on the hardware used for experiments, such as GPU or CPU models. |
| Software Dependencies | No | All experiments are performed on the federated learning framework Plato (Li et al. 2023a). |
| Experiment Setup | Yes | In each communication round, each client does the local training for 5 epochs. We set the upload and download data transmit rate to 100Mbps. Random seeds are fixed during all experiments. |