190 search results - page 7 / 38 » Unsupervised estimation for noisy-channel models |
SDM
2004
SIAM
14 years 11 months ago
2004
SIAM
This paper presents a method of generating Mercer Kernels from an ensemble of probabilistic mixture models, where each mixture model is generated from a Bayesian mixture density e...
98
Voted
NLPRS
2001
Springer
15 years 2 months ago
2001
Springer
We describe a new tagging model where the states of a hidden Markov model (HMM) estimated by unsupervised learning are incorporated as the features in a maximum entropy model. Our...
ICIP
2007
IEEE
15 years 11 months ago
2007
IEEE
Lips segmentation is a very important step in many applications such as automatic speech reading, MPEG-4 compression, special effects, facial analysis and emotion recognition. In ...
114
click to vote
ICMLA
2008
14 years 11 months ago
2008
Time-series segmentation in the fully unsupervised scenario in which the number of segment-types is a priori unknown is a fundamental problem in many applications. We propose a Ba...
NECO
2002
14 years 9 months ago
2002
A Bayesian ensemble learning method is introduced for unsupervised extraction of dynamic processes from noisy data. The data are assumed to be generated by an unknown nonlinear ma...