|
What other approach is similar and can address the same kind of problems as HMM? I think that HMM represents my problem very accurately, but the output is too jumpy to be acceptable, so I am looking into some similar techniques to give them a shot. Thank you very much P.S. Here is where I describe my problem http://metaoptimize.com/qa/questions/9396/what-kind-of-model |
|
CRFs are sort of a generalization of HMMs. You can add more interesting features and do global rather than local normalization.CRFs |
It is hard to name alternatives without knowing what your problem actually is.
Is your problem an instance of changepoint detection / anomaly detection? See for example http://www.cs.toronto.edu/~rpa/papers/rpa-changepoint.pdf for a description of what the problem is.