LCS Book Cover

Design and Analysis of Learning Classifier Systems

A Probabilistic Approach
Series: Studies in Computational Intelligence , Vol. 139
Springer, 2008, 268 pages, ISBN: 978-3-540-79865-1

About the Book

This book provides a comprehensive introduction to the design and analysis of Learning Classifier Systems (LCS) from the perspective of machine learning. LCS are a family of methods for handling unsupervised learning, supervised learning and sequential decision tasks by decomposing larger problem spaces into easy-to-handle subproblems. Contrary to commonly approaching their design and analysis from the viewpoint of evolutionary computation, this book instead promotes a probabilistic model-based approach, based on their defining question "What is an LCS supposed to learn?". Systematically following this approach, it is shown how generic machine learning methods can be applied to design LCS algorithms from the first principles of their underlying probabilistic model, which is in this book -- for illustrative purposes -- closely related to the currently prominent XCS classifier system. The approach is holistic in the sense that the uniform goal-driven design metaphor essentially covers all aspects of LCS and puts them on a solid foundation, in addition to enabling the transfer of the theoretical foundation of the various applied machine learning methods onto LCS. Thus, it not only advances the analysis of existing LCS but also puts forward the design of new LCS within that same framework.

Buy the Book

Directly from Springer
From Amazon US UK DE

Table of Content

  1. Introduction
    1. Machine Learning
    2. Learning Classifier Systems
    3. About the Model-Centred Approach to LCS
    4. How to Read This Book
  2. Background
    1. A General Problem Description
    2. Early Learning Classifier Systems
    3. The LCS Renaissance
    4. Existing Theory
    5. Discussion and Conclusion
  3. A Learning Classifier Systems Model
    1. Task Definition
    2. LCS as Parametric Models
    3. Summary and Outlook
  4. A Probabilistic Model for LCS
    1. The Mixture-of-Experts Model
    2. Expert Models
    3. Generalising the MoE Model
    4. Independent Classifier Training
    5. A Brief Comparison to Linear LCS Models
    6. Discussion and Summary
  5. Training the Classifiers
    1. Linear Classifier Models and Their Underlying Assumptions
    2. Batch Learning Approaches to Regression
    3. Incremental Learning Approaches to Regression
    4. Empirical Demonstration
    5. Classification Models
    6. Discussion and Summary
  6. Mixing Independently Trained Classifiers
    1. Using the Generalised Softmax Function
    2. Heuristic-Based Mixing Models
    3. Empirical Comparison
    4. Relation to Previous Work and Alternatives
    5. Summary and Outlook
  7. The Optimal Set of Classifiers
    1. What is Optimal?
    2. A Fully Bayesian LCS for Regression
    3. Evaluating the Model Evidence
    4. Predictive Distribution
    5. Model Modifications to Perform Classification
    6. Alternative Model Selection Methods
    7. Discussion and Summary
  8. An Algorithmic Description
    1. Computing p(M|D)
    2. Two Alternatives for Model Structure Search
    3. Empirical Demonstration
    4. Improving the Model Structure Search
    5. Summary
  9. Towards Reinforcement Learning with LCS
    1. Problem Definition
    2. Dynamic Programming and Reinforcement Learning
    3. Reinforcement Learning with LCS
    4. Stability of RL with LCS
    5. Further Issues
    6. Summary
  10. Concluding Remarks
  1. Notation
  2. XCS and XCSF
    1. Classifier Model and Mixing Model
    2. Model Structure Search