Loading...
Please wait, while we are loading the content...
Similar Documents
Learning Hidden Markov Models with Geometric Information Learning Hidden Markov Models with Geometric Information
| Content Provider | Semantic Scholar |
|---|---|
| Author | Shatkay, Hagit Kaelbling, Leslie Pack |
| Copyright Year | 1997 |
| Abstract | Hidden Markov models (hmms) and partially observable Markov decision processes (pomdps) provide a useful tool for modeling dynamical systems. They are particularly useful for representing environments such as road networks and ooce buildings, which are typical for robot navigation and planning. In a previous paper SK97] we have empirically shown that by taking advantage of readily available odometric information, learning hmms/pomdps can be made faster and better. This paper extends our previous results in two ways. We show that our algorithm, which introduces an additional data structure and constraints to the standard hmm and extends the Baum-Welch algorithm to accommodate them, indeed converges to a local maximum of the likelihood function. In addition, we extend our empirical study, and show that the algorithm is robust in the face of reduction in the length of the training sequences. Thus, the use of local odometric information yields faster convergence to better solutions with less data. |
| File Format | PDF HTM / HTML |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
