Skip to main content
Disciplines
Automotive Business IT + Informatics Construction + Real Estate Electrical Engineering + Electronics Energy + Sustainability Insurance + Risk Finance + Banking Management + Leadership Marketing + Sales Mechanical Engineering + Materials
Events
DE
EN
Books
Journals
Topic Page
Marketing
Start single access now
Access for companies
EXTENDED SEARCH
Log in
Top

2013 | Book

Introduction Read chapter Read first chapter

Pattern Recognition and Classification

An Introduction

Author: Geoff Dougherty

Publisher: Springer New York

Part of: Springer Professional "Wirtschaft+Technik" , Springer Professional "Technik" , Springer Professional "Wirtschaft"

Login to get access
insite
SEARCH

About this book

The use of pattern recognition and classification is fundamental to many of the automated electronic systems in use today. However, despite the existence of a number of notable books in the field, the subject remains very challenging, especially for the beginner.

Pattern Recognition and Classification presents a comprehensive introduction to the core concepts involved in automated pattern recognition. It is designed to be accessible to newcomers from varied backgrounds, but it will also be useful to researchers and professionals in image and signal processing and analysis, and in computer vision. Fundamental concepts of supervised and unsupervised classification are presented in an informal, rather than axiomatic, treatment so that the reader can quickly acquire the necessary background for applying the concepts to real problems. More advanced topics, such as semi-supervised classification, combining clustering algorithms and relevance feedback are addressed in the later chapters.

This book is suitable for undergraduates and graduates studying pattern recognition and machine learning.

Table of Contents

Frontmatter
Chapter 1. Introduction
Abstract
Pattern recognition involves the recognition of objects or patterns. Classification involves sorting out particular objects into separate distinguishable categories or classes. There are wide varieties of techniques that can be used, and the advent of powerful computers has increased the demand for practical applications. Pattern recognition and classification are at the heart of most machine intelligence systems built for decision making.
Geoff Dougherty
Chapter 2. Classification
Abstract
Classification assigns objects to various classes based on measured features. The features are considered as a feature vector in feature space. It is important to select the most informative features and/or combine features for successful classification. Typically a sample set (the training set) is selected to train the classifier, which is then applied to other objects (the test set). Supervised learning uses a labeled training set, in which it is known to which class the objects belong, and is an inductive reasoning process. There are a variety of approaches to classification; statistical approaches, characterized by an underlying probability model, are very important. We will consider a number of robust features and examples based on shape, size, and topology to classify various objects.
Geoff Dougherty
Chapter 3. Nonmetric Methods
Abstract
There are a number of classification methods for applications involving categorical data [either nominal (unordered categories) or ordinal (ordered categories)], where the objects are described by lists of attributes. A popular method is the decision tree, which uses a branching structure with a series of questions. The questions should be organized so that the most informative are asked first. The information gain is a descriptor of the relative utility of different questions (addressing particular features) at each node of the tree, and it can be formulated in terms of entropy.
Geoff Dougherty
Chapter 4. Statistical Pattern Recognition
Abstract
Probability theory is the basis for understanding and modeling random error. Conditional probability leads to Bayes’ rule and allows us to incorporate measured class-conditional probability density functions into a statistical classifier. The naïve Bayes classifier is a simple probabilistic classifier that assumes independent features. The multivariate Gaussian describes normal distributions in multi-dimensional feature space and is characterized by elliptical isocontours. The covariance matrix summarizes the relationships between the various features. Diagonalization of the covariance matrix by Principal Component Analysis results in independent feature combinations. The Mahalanobis distance is a generalization of the Euclidean distance and can be used to classify objects on the basis of the distance to the class center.
Geoff Dougherty
Chapter 5. Supervised Learning
Abstract
Supervised learning uses a labeled training set of typical objects. The learning refers to some form of adaptation of the classification algorithm to achieve a better response, which will help in classifying the unknown test set. Parametric methods use some form of probability distribution, while non-parametric methods use arbitrary distributions of unknown densities. Parametric learning relies on Bayesian decision theory, and can be easily linked to the concepts of discriminant functions and decision boundaries. [Non-parametric methods include artificial neural networks (ANNs) and support vector machines (SVMs).]
Geoff Dougherty
Chapter 6. Nonparametric Learning
Abstract
With parametric methods, the forms of the underlying density functions are known, and are generally taken as Gaussian. However, these parametric forms do not always fit the probability densities encountered in practice. Most of the classical parametric densities are unimodal, whereas many practical problems involve multimodal densities. For arbitrary distributions of unknown densities, nonparametric methods need to be employed.
Geoff Dougherty
Chapter 7. Feature Extraction and Selection
Abstract
The computational complexity of a classification algorithm should be reduced to a sufficient minimum by reducing the number of features considered. We can either select the most informative features or extract a new, smaller set of features using a (linear) combination of the original features. Principal component analysis (PCA) implements the second option by diagonalizing the covariance matrix to find directions in feature space corresponding to the directions of greatest variance. Linear discriminant analysis (LDA) also reduces the dimensionality of a problem, but specifically finds the most useful directions for separating classes.
Geoff Dougherty
Chapter 8. Unsupervised Learning
Abstract
In unsupervised learning, the class labels are not known, and the data are plotted to see whether it clusters naturally. Clusters can be either partitional or hierarchical. The most common techniques are k-means clustering (for partitional clustering) and agglomerative hierarchical clustering (for hierarchical clustering).
Geoff Dougherty
Chapter 9. Estimating and Comparing Classifiers
Abstract
A variety of metrics have been used to estimate the performance of a classifier, and hence to compare different classifiers. Performance is specific to a particular problem and dataset, and there is no overall best classifier. Cross-validation and resampling methods need to be chosen carefully. Receiver operating characteristic (ROC) curves provide a convenient graphical method of comparing classifier performance, although there are a number of other statistical tests.
Geoff Dougherty
Chapter 10. Projects
Abstract
Pattern recognition is a rich field for project work. It is useful in many applications such as information retrieval, data mining, document image analysis and recognition, computational linguistics, forensics, biometrics and bioinformatics. A few possible applications of pattern recognition/classification techniques are demonstrated in this chapter by reference to a few specific projects, some in more detail than others.
Geoff Dougherty
Backmatter
Metadata
Title
Pattern Recognition and Classification
Author
Geoff Dougherty
Copyright Year
2013
Publisher
Springer New York
Electronic ISBN
978-1-4614-5323-9
Print ISBN
978-1-4614-5322-2
DOI
https://doi.org/10.1007/978-1-4614-5323-9

Premium Partner

    Image Credits