Anti-likelihood를 이용한 HMM 구조의 최적화 = HMM Topology Optimization Using Anti-likelihood

다국어 초록 (Multilingual Abstract)

HMMs are frequently used in on-line handwriting recognition and their topologies is decidedby the number of states, the number of mixtures per state and the transition probabilities. It is very difficult to select the optimum model topology, since HMM topology is decided by some heuristic methods in most of previous researches. In this paper, we proposed an anti-likelihood criterion for systematic HMM topology optimization method that improves discrimination power between models. We also performed some comparative research on couple of model selection criteria for online handwriting data recognition. We got better recognition results with fewer number of parameters.