1. Introduction

Genetics-based Machine Learning (GBML) is the application of Evolutionary Algorithms (EAs) to machine learning. We assume readers are familiar with EAs, which are well documented elsewhere, and their application to optimisation problems. In this introductory section we outline the scope of machine learning, introduce the more specific area of supervised learning, and contrast it with optimisation. However, the treatment is necessarily brief and readers who desire to work in GBML are strongly advised to first gain a solid foundation in non-evolutionary approaches to machine learning. Section §2 describes a framework for GBML which includes ways of classifying GBML algorithms and a discussion of the interaction between learning and evolution. Section §3 reviews the work of a number of GBML communities with emphasis on their evolutionary aspects. Finally, section §4 concludes.

What's Missing

Given the breadth of the field and the volume of the literature the coverage herein is necessarily somewhat arbitrary and misses a number of significant subjects. These include a general introduction to machine learning including the structure of learning problems and their fitness landscapes (which we must exploit in order to learn efficiently), non-evolutionary algorithms (which constitute the majority of machine learning methods, and include both simple and effective methods), and theoretical limitations of learning (such as the no free lunch theorem for supervised learning [311] and the conservation law of generalisation [245]). Also missing is coverage of GBML for clustering, reinforcement learning, Bayesian networks, artificial immune systems, artificial life, and application areas. Finally, some areas which have been touched on have been given an undeservedly cursory treatment, including EAs for data preparation (e.g. feature selection), co-evolution, and comparisons between GBML and non-evolutionary alternatives. However, [95] contains good treatments of GBML for, among others, clustering and data preparation.

Subsections

1.1 Machine Learning

• Inductive Generalisation
• 1-Max: a Typical Optimisation Problem
• Classification of Mushrooms: a Typical Learning Problem
• Terminology in Supervised Learning
• Comparison of Supervised Learning and Optimisation
• Issues in Supervised Learning

1.2 Arguments For and Against GBML