Self-Supervised Graph Attention Networks for Deep Weighted Multi-View Clustering
Authors: Zongmo Huang, Yazhou Ren, Xiaorong Pu, Shudong Huang, Zenglin Xu, Lifang He
AAAI 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experimental results on multiple real-world datasets demonstrate the effectiveness of our method over existing approaches. |
| Researcher Affiliation | Academia | 1School of Computer Science and Engineering, University of Electronic Science and Technology of China, Chengdu, China 2 Shenzhen Institute for Advanced Study, University of Electronic Science and Technology of China, Shenzhen, China 3College of Computer Science, Sichuan University, Chengdu, China 4School of Computer Science and Technology, Harbin Institute of Technology Shenzhen, Shenzhen, China 5Department of Computer Science and Engineering, Lehigh Univerisity, Bethlehem, USA |
| Pseudocode | Yes | Algorithm 1: The SGDMC model. |
| Open Source Code | No | The paper does not explicitly state that the source code for the described methodology is publicly available, nor does it provide a link to a code repository. |
| Open Datasets | Yes | Three widely used and publicly available multi-view datasets are implemented in our study: BDGP (Cai et al. 2012) consists of 2500 samples of 5 different kinds of drosophila embryos. ... Handwritten Numerals1 sources from UCI machine learning repository, which contains 2000 handwritten numeral images over 10 classes (0-9). Each instance has six visual views, including 216 profile correlations, 76 Fourier coefficients of the character shapes, 64 Karhunen-Love coefficients, 6 morphological features, 240 pixel averages in 2 3 windows, and 47 Zernike moments. Reuters2 is comprised of 1200 articles in 6 categories (C15, CCAT, E21, ECAT, GCAT and M11), each providing 200 articles. For each article, it is written in five different languages (English, French, German, Italian, and Spanish). |
| Dataset Splits | No | The paper does not explicitly provide training/validation/test dataset splits. It mentions that 'All feature learning encoders are pretrained for 2000 epochs' and 'The training process compulsively terminates when the epoch number exceeds Tmax = 10000', but no details on data partitioning into distinct sets for training, validation, and testing. |
| Hardware Specification | No | The paper does not provide specific hardware details (exact GPU/CPU models, processor types with speeds, memory amounts, or detailed computer specifications) used for running its experiments. |
| Software Dependencies | No | The paper does not provide specific ancillary software details (e.g., library or solver names with version numbers) needed to replicate the experiment. |
| Experiment Setup | Yes | Specifically, for each view, the structure of encoder is: Input(dv) Fc500 Fc500 Fc2000 Fc10(d v), and the decoder is symmetric with the encoder. All feature learning encoders are pretrained for 2000 epochs. The aligned rate threshold δ is 0.8 and the number of neighbors β applied to construct the adjacent graph is set to 10. The batch size is set to the instance number n. During the finetuning stage, the pseudo-label P and the sample weights W v are updated for every T = 1000 epochs. The training process compulsively terminates when the epoch number exceeds Tmax = 10000. |