ITopology Interface

An Hidden Markov Model Topology specifies how many states and which initial probabilities a Markov model should have. Two common topologies can be discussed in terms of transition state probabilities and are available to construction through the Ergodic and Forward classes implementing this interface.

Topology specification is important with regard to both learning and performance: A model with too many states (and thus too many settable parameters) will require too much training data while an model with an insufficient number of states will prohibit the HMM from capturing subtle statistical patterns.

References:

• Alexander Schliep, "Learning Hidden Markov Model Topology".
• Richard Hughey and Anders Krogh, "Hidden Markov models for sequence analysis: extension and analysis of the basic method", CABIOS 12(2):95-107, 1996. Available in: http://compbio.soe.ucsc.edu/html_format_papers/hughkrogh96/cabios.html

Reference