Confusion-Resistant Federated Learning via Diffusion-Based Data Harmonization on Non-IID Data
Authors: xiaohong chen, Canran Xiao, Yongmei liu
NeurIPS 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments on benchmark datasets, including MNIST, Fashion MNIST, CIFAR-10, CIFAR-100, and NIPD, demonstrate the effectiveness of CRFed in improving accuracy, convergence speed, and overall robustness in federated learning scenarios with severe data heterogeneity. |
| Researcher Affiliation | Academia | Xiaohong Chen1,2,3 Canran Xiao1 Yongmei Liu1,4 1 School of Business, Central South University, Changsha, Hunan 410083, China 2 Xiangjiang Laboratory, Changsha, Hunan 410205, China 3 School of Advanced Interdisciplinary Studies, School of Management Science and Engineering, Hunan University of Technology and Business, Changsha, Hunan 410205, China 4 Urban Smart Governance Laboratory, Changsha, Hunan 410083, China |
| Pseudocode | Yes | The complete computational process(pseudocode) of CRFed is provided in the A.5. (Algorithm 1 Confusion-Resistant Federated Learning via Consistent Diffusion (CRFed)) |
| Open Source Code | Yes | Question: Does the paper provide open access to the data and code, with sufficient instructions to faithfully reproduce the main experimental results, as described in supplemental material? Answer: [Yes] Justification: The data and code is provided. |
| Open Datasets | Yes | Our experiments are conducted on four widely used benchmark datasets: MNIST [Le Cun et al., 1998], Fashion MNIST [Xiao et al., 2017], CIFAR10 [Krizhevsky et al., 2009], and CIFAR100 [Krizhevsky et al., 2009]. ...Additionally, we use the NIPD dataset [Yin et al., 2023], a benchmark specifically designed for federated learning in person detection tasks with Non-IID data. |
| Dataset Splits | No | The paper describes how data is partitioned among clients for non-IID scenarios using the Dirichlet distribution and specifies data samples per node (e.g., "each node has 600 data samples" for MNIST). However, it does not provide explicit overall training, validation, and test dataset splits (e.g., 80/10/10 percentages or specific counts for a global dataset split). |
| Hardware Specification | Yes | The experiments were conducted using an NVIDIA Ge Force RTX 4060 GPU, which has 8GB of VRAM. |
| Software Dependencies | No | The paper mentions "Momentum optimization with a coefficient of 0.5 is applied" but does not specify any software libraries or frameworks with version numbers (e.g., PyTorch, TensorFlow, or scikit-learn versions). |
| Experiment Setup | Yes | For local training, the settings are as follows: MNIST with E = 5, B = 10, η = 5 10 3; Fashion MNIST with E = 5, B = 100, η = 2 10 4; CIFAR-10 and CIFAR100 with E = 5, B = 100, η = 1 10 4. Momentum optimization with a coefficient of 0.5 is applied. In the CRFed framework, key hyperparameters include maximum global rounds (TG) set to 100, local training cycles (El) per global round set to 1, regularization coefficient (λ) set to 0.1, dynamically adjusted confidence threshold (τ), and client selection threshold (γ) initially set to 0.5. |