Reproduce “Semi-Supervised Clustering via Dynamic Graph Structure Learning” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Huaming Ling, Chenglong Bao, Xin Liang, Zuoqiang Shi
The Dynamic Graph Structure Learning (DGSL) framework concurrently optimizes the affinity matrix and low-dimensional data representations for semi-supervised clustering. This approach, leveraging both local and global data relationships, consistently outperforms baseline methods across various benchmark datasets.
Key insights
- Optimal clustering performance necessitates jointly learning both data representations and the underlying graph structures, allowing them to mutually reinforce each other.
- Dynamically updating the graph structure based on evolving data representations (rather than keeping it static) enhances the adaptability and effectiveness of clustering.
- Integrating both local Euclidean distance measures and global self-representation models in the affinity matrix construction captures a more comprehensive understanding of data relationships.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.