Machine Learning

Chapter 11

Hidden Markov Model

11.1 Introduction

A hidden Markov model (HMM) has invisible or unobservable states, but a visit to these hidden states results in the recording of observation that is a probabilistic function of the state. Given a sequence of observations, the problem is to infer the dynamical system that had produced the given sequence. This inference will result in a model for the underlying process. The three basic problems of HMMs are given below:

1. Scoring problem: The target is to find the probability of an observed sequence with HMM already given.

2. Alignment problem: Given a particular HMM, the target is to determine from an observation sequence the most likely sequence of underlying hidden states that might have generated ...