ML Anthology

SMEM Algorithm Is Not Fully Compatible with Maximum-Likelihood Framework

Akihiro Minagawa, Norio Tagawa, Toshiyuki Tanaka
NeCo 2002 pp. 1261-1266
doi:10.1162/089976602753712927 /neco/2002/minagawa2002neco-smem/

Abstract

The expectation-maximization (EM) algorithm with split-and-merge operations (SMEM algorithm) proposed by Ueda, Nakano, Ghahramani, and Hinton (2000) is a nonlocal searching method, applicable to mixture models, for relaxing the local optimum property of the EM algorithm. In this article, we point out that the SMEM algorithm uses the acceptance-rejection evaluation method, which may pick up a distribution with smaller likelihood, and demonstrate that an increase in likelihood can then be guaranteed only by comparing log likelihoods.

NeCo Semantic Scholar

Cite

Text

Minagawa et al. "SMEM Algorithm Is Not Fully Compatible with Maximum-Likelihood Framework." Neural Computation, 2002. doi:10.1162/089976602753712927

Markdown

[Minagawa et al. "SMEM Algorithm Is Not Fully Compatible with Maximum-Likelihood Framework." Neural Computation, 2002.](https://mlanthology.org/neco/2002/minagawa2002neco-smem/) doi:10.1162/089976602753712927

BibTeX

@article{minagawa2002neco-smem,
  title     = {{SMEM Algorithm Is Not Fully Compatible with Maximum-Likelihood Framework}},
  author    = {Minagawa, Akihiro and Tagawa, Norio and Tanaka, Toshiyuki},
  journal   = {Neural Computation},
  year      = {2002},
  pages     = {1261-1266},
  doi       = {10.1162/089976602753712927},
  volume    = {14},
  url       = {https://mlanthology.org/neco/2002/minagawa2002neco-smem/}
}