(ML 14.6) Forward-Backward algorithm for HMMs

[raven112345]

Why do we need to assume the emmission probabilities, transitional probabilities and the initialization probabilities?

[pelemanov]

@raven112345 You need them to be able to do the inference. When you want to use this in practice, you will typically build your own HMM and define these yourself. Then you will use this algorithm to inference what you don't know.

[pelemanov]

That should be "infer", not "inference" the second time.

[MrLufemami]

Could anyone tell me how to distinguish different numerical sequences just by looking at a number that corresponds to a specific sequence that is different from another sequence...like in a classifer of sequences...have tried entropy but it gives a lot of false positives in the identification of the sequences