Overview
Baibo Zhang received his B.E. in Automation Technology from Tsinghua University (1999) and was a Ph.D. candidate in the Department of Automation, Tsinghua University at the time of publication. His research focuses on unsupervised learning, pattern recognition, and handwritten Chinese character analysis. He is the first author of the Competitive EM (CEM) algorithm paper.
Key Contributions / Features
- Zhang-Zhang-Yi (2004): Co-developed the CEM algorithm for finite mixture models — competitive split/merge via local Kullback divergence probabilities J(m;θ), MML-based criterion, Metropolis acceptance, CEM² sequential annihilation. CEM achieves Ps = 90.2–96.9% on unequal-prior mixtures vs. AEM's 60.3–71.0% and EM's 19.0–59.2%. Application to Chinese character stroke extraction. See Zhang-Zhang-Yi (2004).
Related