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 Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..
DKDR: Dynamic Knowledge Distillation for Reliability in Federated Learning
Authors: Yueyang Yuan, Wenke Huang, Frank Wan, Kaiqi Guan, He Li, Mang Ye
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Empirical results from single-domain and multi-domain image classification tasks demonstrate the effectiveness of the proposed method and the efficiency of its key modules. The code is available at https://github.com/Yueyang Yuan/DKDR. 4 Experiments 4.1 Experimental Setup Datasets. We evaluate DKDR on two single-domain scenarios and two multi-domain scenarios. Table 2: Comparison with the state-of-the-art method in the Office31 and Office Home with domain skew. Best in bold and second with underline. Please refer to Sec.4.2 for further explanations. |
| Researcher Affiliation | Academia | Yueyang Yuan1,2 , Wenke Huang1,2 , Guancheng Wan1 , Kaiqi Guan1, He Li1, Mang Ye1 1 National Engineering Research Center for Multimedia Software, Institute of Artificial Intelligence, Hubei Key Laboratory of Multimedia and Network Communication Engineering, School of Computer Science, Wuhan University, Wuhan, China. 2 Guangdong Laboratory of Artificial Intelligence and Digital Economy (SZ) EMAIL |
| Pseudocode | No | The paper describes its methodology in narrative text and mathematical equations in Section 3, and illustrates its architecture in Figure 3, but does not present any explicitly labeled pseudocode or algorithm blocks with structured steps. |
| Open Source Code | Yes | Empirical results from single-domain and multi-domain image classification tasks demonstrate the effectiveness of the proposed method and the efficiency of its key modules. The code is available at https://github.com/Yueyang Yuan/DKDR. |
| Open Datasets | Yes | Datasets. We evaluate DKDR on two single-domain scenarios and two multi-domain scenarios. Cifar-10 [23] contains 50k training images and 10k test images with 32 32 for 10 classes. Cifar-100 [23] contains 50k and 10k images with 32 32 for 100 classes. Office31 [44] consists of three domains: Amazon (A), Webcam (W) and DSLR (D). Office Home [51] consists of four domains: Art (A), Clipart (C), Product (P), and Real World (R). |
| Dataset Splits | Yes | Cifar-10 [23] contains 50k training images and 10k test images with 32 32 for 10 classes. Cifar-100 [23] contains 50k and 10k images with 32 32 for 100 classes. |
| Hardware Specification | Yes | We conduct experiments with Res Net-10 [12] on single-domain scenarios and Res Net-18 [12] on multi-domain scenarios. We fix the random seed to ensure reproduction and conduct experiments on the NVIDIA 3090Ti. |
| Software Dependencies | No | We use the SGD optimizer with the learning rate lr = 1e 3. The corresponding weight decay is 1e 5 and momentum is 0.9. The training batch size is 64 for single-domain tasks and 16 for multi-domain tasks. |
| Experiment Setup | Yes | Implement Details. We conduct communication epoch for E = 200 and local updating round T = 5, where all federated learning approaches have little or no accuracy gain with more communications. We use the SGD optimizer with the learning rate lr = 1e 3. The corresponding weight decay is 1e 5 and momentum is 0.9. The training batch size is 64 for single-domain tasks and 16 for multi-domain tasks. The client number K is 20 for different datasets. We conduct experiments with Res Net-10 [12] on single-domain scenarios and Res Net-18 [12] on multi-domain scenarios. We fix the random seed to ensure reproduction and conduct experiments on the NVIDIA 3090Ti. |