README.md

Divide-and-conquer based Large-Scale Spectral Clustering

You can find our paper in ArXiv.org Or ResearchGate.

Datasets

Five real world datasets and five synthetic datasets:

• PenDigits
• USPS
• Letters
• MNIST
• Covertype
• Three Spirals-60K (TS-60K)
• Two Moons-1M (TM-1M)
• Three Circles-6M (TC-6M)
• Circles and Gaussian-10M (CG-10M)
• Flower-20M (FL-20M)

Algorithm

• Divide-and-conquer based Large-scale Spectral Clustering

• Divide-and-conquer based Landmark Selection
• Approximate K-nearest landmarks method (line 35~62 of DnC_SC)

Code

See our demo.

Reference

If you find this code useful for your research, please cite

@misc{li2021divideandconquer,
    title={Divide-and-conquer based Large-Scale Spectral Clustering},
    author={Hongmin Li and Xiucai Ye and Akira Imakura and Tetsuya Sakurai},
    year={2021},
    eprint={2104.15042},
    archivePrefix={arXiv},
    primaryClass={cs.LG}
}