Data Classification

Chapter 20

Semi-Supervised Learning

Kaushik Sinha

Wichita State UniversityWichita, KS [e-mail]

20.1 Introduction

Consider an input space X and an output space Y, where one would like to see an example from input space X and automatically predict its output. In traditional supervised learning, a learning algorithm is typically given a training set of the form ${\left\{\left(x_i,y_i\right)\right\}}_{i=1,}^l$ where each pair (xi, yi) ∈ X × Y is drawn independently at random according to an unknown joint probability distribution PX×Y .In the case of a supervised classification problem, Y is a finite set of class labels and the goal of the learning algorithm is to construct a function g : X → Y that predicts the label y given x. For example, consider the problem ...