Feature Extraction and Selection

Abstract

One of the tasks of pattern recognition is to convert patterns to features, where these features are a description of the collected data in a compact form. Ideally, these features only contain relevant information, which then play a crucial role in determining the division of properties concerning each class. Mathematical models of feature extraction lead to a dimensionality reduction, resulting in lower-dimensional representation of the information. Following feature extraction, feature selection has an important influence on classification accuracy, necessary time for classification, the number of examples for learning, and the cost of performing classification.