Scalable One-Pass Incomplete Multi-View Clustering by Aligning Anchors
Authors: Yalan Qin, Guorui Feng, Xinpeng Zhang
AAAI 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experiments on several incomplete multi-view datasets validate the efficiency and effectiveness of SOME-AS. |
| Researcher Affiliation | Academia | Yalan Qin, Guorui Feng *, Xinpeng Zhang *, School of Communication and Information Engineering, Shanghai University, Shanghai, China EMAIL |
| Pseudocode | Yes | Algorithm 1: Algorithm of SOME-AS |
| Open Source Code | No | The paper does not provide explicit statements about releasing source code or links to a code repository. |
| Open Datasets | Yes | We adopt seven widely used datasets in the evaluation. ORL consists of 400 face images and 3 views, whose dimensions are 3304, 4096 and 6750, respectively. BDGP contains 2500 samples from 5 categories. Protein Fold is a protein dataset consisting of 27 classes, whose data size is 2500. SUNRGBD is the dataset containing 10335 samples from 45 categories. NUSWIDEOBJ consists of 30000 samples from 31 classes. Cifar10 contains 50000 images from 10 categories. MNIST is the digit dataset consisting of 60000 samples. |
| Dataset Splits | No | The paper mentions randomly removing samples to create incomplete multi-view datasets and testing with different 'missing rates' (10% to 90%). However, it does not provide specific training/test/validation dataset splits or their percentages, counts, or methodologies for model evaluation in the traditional supervised learning sense. |
| Hardware Specification | Yes | We perform all experiments on a computer with AMD Ryzen 5 1600X Six-Core Processor. |
| Software Dependencies | No | The paper does not specify any software dependencies with version numbers (e.g., Python, PyTorch, TensorFlow versions). |
| Experiment Setup | Yes | There are total two parameters needed to be determined in the experiment, i.e., α and β, where α is the parameter of Frobenius norm and β corresponds to the alignment mapping parameter. We select them in the range of [0.001, 0.01, 0.1, 1, 10] on dataset. According to Figs. 2-3, we find that satisfied results can be achieve when α and β are both 0.1. Moreover, it is observed that the clustering performance is relatively stable in the given range of parameters on dataset. For the parameters of the compared methods, we set them as the recommended ones in the corresponding literatures. |