Machine Learning for the Web

Hidden Markov model

Although this method cannot be strictly considered a supervised learning algorithm, it can be also used to perform something that is really similar to classification, so we decided to include it here. To introduce the subject, we are going to present an example. Consider the simplistic case of predicting whether a salesman in front of you is lying or not (two states ) by observing his glance: eye contact, looking down, or looking aside (each observation O_i has the values 0, 1, and 2 respectively). Imagine a sequence of observations of the salesman's glances O=O₀, O₁, O₂, O₃, O₄,… are 0, 1, 0, 2,… We want to infer the transition ...