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].

Data-Free Diversity-Based Ensemble Selection for One-Shot Federated Learning

Authors: Naibo Wang, Wenjie Feng, yuchen deng, Moming Duan, Fusheng Liu, See-Kiong Ng

TMLR 2023 | Venue PDF | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental results show that our method can achieve both better performance and higher efficiency over 7 datasets, 5 different model structures, and both homogeneous and heterogeneous model groups under four different data-partition strategies.
Researcher Affiliation Academia 1 Institute of Data Science, National University of Singapore 2 School of Mathematics and Statistics, Changchun University of Technology
Pseudocode Yes Algorithm 1: De DES framework Algorithm 2: Outlier Filter algorithm for the model filtering
Open Source Code No The paper does not provide concrete access to source code. It does not contain an explicit statement about code release, a link to a code repository, or mention of code in supplementary materials.
Open Datasets Yes We used 7 image datasets and 5 types of neural network models in our experiments, details can be found in the Appendix. [...] Figure 3 shows an example of the data distribution under the different partition strategies for CIFAR-10 with 5 parties. [...] EMNIST Digits [...] EMNIST Balanced [...] SVHN [...] FEMNIST [...] CIFAR10 [...] CIFAR100
Dataset Splits Yes To simulate the real scenarios in FL (Li et al., 2022) we designed four types of dataset-partition strategies to evaluate De DES, which lead to different local data distribution to train diverse client models Mi. Homogeneous (homo): the amount of samples and the data distribution keep the same for all parties. IID but different quantity (iid-dq): the training data of each party follows the same distribution, but the amount of data is different. Skewed data distribution (noniid-lds): the training data of each party follows different distributions, especially for the label distribution. Non-iid with k (< C) classes (noniid-lk): the training data of each party only contains k of C classes, which is an extreme non-iid setting.
Hardware Specification No The paper does not provide specific hardware details (e.g., exact GPU/CPU models, processor types with speeds, memory amounts, or detailed computer specifications) used for running its experiments. It mentions training models but not on what hardware.
Software Dependencies No The paper does not provide specific ancillary software details, such as library or solver names with version numbers, needed to replicate the experiment.
Experiment Setup No The detailed runtime setups and configuration of De DES are elaborated in the Appendix, including the learning rate, model representation strategy, clustering method for different data partitions, etc.