Could you please detail on your comment that rounding errors may occur with Gaussian elimination / LU factorization?

I would be happy to better understand this point.

Thanks

Sebastian

and you should really use MathJax because obviously! :)

Also, you should have mentioned that the BW algorithm will not give you a global maximum (max probability for observing a particular sequence), it gives only a local maximum. But one can alleviate this problem by having many initial randomly distributed initial probabilities. The things get a little bit more interesting when we move toward the continuous state space hidden markovian processes. ;)

Anyhow, I love your site and the BW algorithm is simply awesome! Have you ever used it in real life applications? if yes, u should also help me get a job in US ^_^

]]>