Federated Partial Label Learning with Local-Adaptive Augmentation and Regularization
Authors: Yan Yan, Yuhong Guo
AAAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments conducted on multiple datasets under the Fed PLL setting demonstrate the effectiveness of the proposed Fed PLLLAAR method for federated partial label learning. |
| Researcher Affiliation | Academia | Yan Yan1, Yuhong Guo1, 2 1School of Computer Science, Carleton University, Ottawa, Canada 2Canada CIFAR AI Chair, Amii yanyan@cunet.carleton.ca, yuhong.guo@carleton.ca |
| Pseudocode | No | No structured pseudocode or algorithm blocks were found in the paper. |
| Open Source Code | No | The paper does not provide any explicit statements or links indicating that open-source code for the described methodology is available. |
| Open Datasets | Yes | Datasets We conducted experiments on three benchmark datasets: (1) CIFAR-10 (Krizhevsky, Hinton et al. 2009)... (2) CIFAR-100 (Krizhevsky, Hinton et al. 2009)... (3) SVHN (Netzer et al. 2011)... |
| Dataset Splits | No | The paper states training parameters like 'The model is trained for 200 communication rounds, with 5, 10, and 10 local epochs in each round on CIFAR-10, SVHN, and CIFAR-100 respectively,' but does not provide specific details about training/validation/test dataset splits (e.g., percentages or sample counts for each split, or mention of a validation set). |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., CPU, GPU models, or memory) used for running the experiments. |
| Software Dependencies | No | The paper mentions 'a stochastic gradient descent optimizer' but does not list specific software libraries or their version numbers (e.g., Python, PyTorch, TensorFlow, CUDA versions) needed to replicate the experiment environment. |
| Experiment Setup | Yes | In the experiment, we distribute the Fed PLL dataset among 10 clients... The model is trained for 200 communication rounds, with 5, 10, and 10 local epochs in each round on CIFAR-10, SVHN, and CIFAR-100 respectively. We adopt Res Net-18 as the backbone network on CIFAR-10 and SVHN, and adopt Res Net-34 on CIFAR-100. ... The moving average hyperparameter µ for pseudo-label updating and the parameter α used for Beta distribution are set to 0.99 and 0.75 respectively. The probability σ and the parameter β used for producing non-IID data are set to 0.7 and 0.5 respectively. For optimization, we adopt a stochastic gradient descent optimizer with a momentum of 0.5. The learning rate and mini-batch size are set to 0.03 and 128 respectively in all our comparison experiments. The hyperparameters η and δ used in Eq.(14) are set to 1 by default. |