EGNMF

EGNMF is a GNMF-based clustering algorithm using cluster ensembles. The algorithm stably achieves a high and robust clustering performance.

Main Results

The following results are obtained from simulation.ipynb which uses the digits dataset.

• Accuracy (AC)

• Normalized Mutual Information (NMI)

Dependencies

• numpy
• scipy
• scikit-learn
• munkres
• pyitlib
• matplotlib
• cluster-ensembles

References

[1] https://doi.org/10.11517/pjsai.JSAI2020.0_4J2GS202

[2] D. Cai, X. He, J. Han, and T. S. Huang, Graph regularized nonnegative matrix factorization for data representation, IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 33, no. 8, pp. 1548–1560, 2010.

[3] X. Z. Fern and C. E. Brodley, Solving cluster ensemble problems by bipartite graph partitioning, In Proceedings of the 21st International Conference on Machine Learning, p. 36, 2004.