FedMSGL: A Self-Expressive Hypergraph Based Federated Multi-View Learning
Authors: Daoyuan Li, Zuyuan Yang, Shengli Xie
AAAI 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experiments on multi-view datasets with different feature dimensions validated the effectiveness of the proposed method. |
| Researcher Affiliation | Academia | 1School of Automation, Guangdong Provincial Key Laboratory of Intelligent Systems and Optimization Integration, Guangdong University of Technology 2Key Laboratory of i Detection and Manufacturing-Io T, Ministry of Education |
| Pseudocode | Yes | Algorithm 1: Fed MSGL Input: Multi-view dataset with n view X1, X2, , Xn, Hyperparameter λ1,λ2 and λ3. Ouput: Global model G and the cluster indicator matrix F. Initialize Ck, Uk 1: while converge do Local node: do calculation among nodes parallel 2: for i = 1 to itermax do 3: if i = 1 then replace Ck with G; 4: else update Ck by Eq.(17); 5: update Uk by Eq.(20); 6: end if 7: end for send Ck and Uk to server; Central Server: 8: for i = 1 to itermax do update G by Eq.(24); 9: update F by solving Eq.(25); 10: end for send G to nodes; 11: end while |
| Open Source Code | No | The paper does not contain any explicit statements about releasing source code, nor does it provide any links to a code repository. |
| Open Datasets | Yes | We conducted experiments on five multi-view datasets and four have differing feature dimensions. Sonar1,BBC Sport2,ORL3 ,Handwritten4 and Caltech101-75. The statistics of the five datasets are summarized in Table.2. ... 1http://archive.ics.uci.edu/dataset/151/connectionist+bench +sonar+mines+vs+rocks 2http://mlg.ucd.ie/datasets/bbc.html 3https://www.cl.cam.ac.uk/research/dtg/attarchive/facedat abase.html 4https://archive.ics.uci.edu/dataset/72/multiple+features 5http://www.vision.caltech.edu/archive.html |
| Dataset Splits | No | The paper lists the datasets used and their statistics (sample, view, feature dimension) in Table 2, but it does not specify any training, testing, or validation splits for these datasets. |
| Hardware Specification | Yes | All experiments are conducted on the i9-14900KF and 32.0GB RAM, MATLAB R2021b. |
| Software Dependencies | Yes | All experiments are conducted on the i9-14900KF and 32.0GB RAM, MATLAB R2021b. |
| Experiment Setup | Yes | Parameters setting For the compared methods mentioned above, we use the parameters recommended by the authors. For the proposed method Fed MSGL, we select λ1, λ2 and λ3 from the range {1e 3, 1e 2, , 1e2, 1e3}. The value of β is chosen from the range {1e 2, 1e 1, 1, 10, 100}. After that, we choose specific hyperparameter combinations based on model performance variations. |