Notice: This website is continuously updated as new conference proceedings and journal volumes are published. Consequently, the figures presented here may differ from those reported in our publications.
A brief overview of our methodology is provided on the Methods page, along with references to publications documenting our approach and links to the code and data used in each study.
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.
FedVS: Straggler-Resilient and Privacy-Preserving Vertical Federated Learning for Split Models
Authors: Songze Li, Duanyi Yao, Jin Liu
ICML 2023 | Venue PDF | LLM Run Details | Input Tokens: 24,544 Total number of tokens sent to the LLM as input for this paper's analysis. | Output Tokens: 4,576 Total number of tokens produced by the LLM (including reasoning/thinking tokens) for this paper's analysis.
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments on various types of VFL datasets (including tabular, CV, and multi-view) demonstrate the universal advantages of Fed VS in straggler mitigation and privacy protection over baseline protocols. |
| Researcher Affiliation | Academia | 1The Hong Kong University of Science and Technology (Guangzhou) 2The Hong Kong University of Science and Technology. Correspondence to: Songze Li <EMAIL>. |
| Pseudocode | Yes | The paper includes “Algorithm 1 The Fed VS protocol” with a structured block of steps. |
| Open Source Code | No | The paper does not provide any statement about releasing open-source code or a link to a code repository. |
| Open Datasets | Yes | We carry out split VFL experiments on three types of six real-world datasets, including Fashion MNIST dataset (Xiao et al., 2017a), Parkinson (Sakar et al., 2019), Credit card (Yeh & Lien, 2009), EMNIST (Cohen et al., 2017), Handwritten (Dua & Graff, 2017) and Caltech-7 (Li et al., 2022). |
| Dataset Splits | Yes | For Parkinson: “70% of the data is regarded as training data and the remaining part is test data.” For Hand Written: “The dataset is split to 60% as the train set and 40% as the test set.” For Caltech-7: “80% of the dataset is used for training and 20% for testing.” |
| Hardware Specification | Yes | All experiments are performed on a single machine using four NVIDIA Ge Force RTX 3090 GPUs. |
| Software Dependencies | No | The paper does not list specific software dependencies with version numbers (e.g., Python 3.x, PyTorch 1.x) used for running its experiments. |
| Experiment Setup | Yes | We provide descriptions of the datasets, number of clients considered for each dataset, employed model architectures, and training parameters in Appendix C. For Parkinson: “The learning rate is set to 0.005. The batch size is 16.” For Credit Card: “The learning rate is set to 0.01 and the batch size is set to 32.” |