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..
Decentralized Dynamic Cooperation of Personalized Models for Federated Continual Learning
Authors: Danni Yang, Zhikang Chen, Sen Cui, Mengyue Yang, Ding Li, Abudukelimu Wuerkaixi, Haoxuan Li, Jinke Ren, Mingming Gong
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | 5 Experiments |
| Researcher Affiliation | Academia | 1Tsinghua University 2Peking University 3University of Bristol 4The Chinese University of Hong Kong, Shenzhen 5The University of Melbourne 6Mohamed bin Zayed University of Artificial Intelligence (MBZUAI) |
| Pseudocode | Yes | Algorithm 1: Merge-Blocking Algorithm |
| Open Source Code | Yes | Code is available at: https://github.com/ydn3229/DCFCL. |
| Open Datasets | Yes | We conduct 4 datasets on different settings. 1) EMNIST-LTP [3]: a character classification dataset with 26 classes. ... 3) CIFAR100 [40]: a challenging image classification datase. 4) MNIST-SVHN-F [9, 41, 42]: The dataset is constructed with MNIST [9], SVHN [41] and Fashion MNIST [42]. |
| Dataset Splits | No | Each client has a local dataset Dk = {D1 k, D2 k, . . . , DT k }, where T denotes the total number of task phases and Dt k = {xti k , yti k }nt k i=1 is the training data in phase t containing nt k samples... Suppose ai,t k is the test set accuracy of the t-th task after learning the i-th task in client k. |
| Hardware Specification | Yes | In the experiments, we conduct all methods on a local Linux server that has two physical CPU chips (Intel(R) Xeon(R) CPU E5-2640 v4 @ 2.40GHz) and 32 logical kernels. All methods are implemented using Pytorch framework and all models are trained on Ge Force RTX 2080 Ti GPUs. |
| Software Dependencies | No | All methods are implemented using Pytorch framework and all models are trained on Ge Force RTX 2080 Ti GPUs. |
| Experiment Setup | Yes | The Adam optimizer is employed for training all models. For all experiments except for CIFAR100, a learning rate of 1e-4 is utilized, with a global communication round of 60, and local iteration of 100. We set learning rate as 1e-3, the global communication round as 40, and local iteration as 400 for CIFAR100. Other parameters include weightdecay = 1e 5, beta1 = 0.9, beta2 = 0.999. For training, a mini-batch size of 64 is adopted. |