Davoian, Kristina: Advancing evolution of artifcial neural networks through behavioral adaptation. 2011
Content
- Abstract
- Acknowledgements
- List of Figures
- List of Tables
- Abbreviations
- 1 Introduction
- 2 Fundamental Concepts
- 2.1 Evolutionary Algorithms
- 2.1.1 Genetic Algorithms
- 2.1.2 Evolution Strategies
- 2.1.3 Evolutionary Programming
- 2.1.4 Genetic Programming
- 2.1.5 Global Convergence and Computational Complexity of EAs
- 2.1.6 Parallelization of EAs: Parallel Evolutionary Algorithms
- 2.2 Artificial Neural Networks
- 2.2.1 Classification of ANNs
- 2.2.2 Learning in ANNs
- 2.2.3 Overview of Supervised Learning Algorithms
- 2.2.4 Generalization in ANNs
- 2.3 Evolutionary Artificial Neural Networks
- 3 Mutation-based Evolutionary Algorithms
- 3.1 Adaptation and Self-Adaptation in EP and ES
- 3.2 Classical Evolutionary Programming
- 3.3 Fast Evolutionary Programming
- 3.4 Combined Approaches
- 3.5 Conclusions
- 4 Including Phenotype Information in Mutation
- 4.1 Genotype Information in NWEA
- 4.2 Deriving Phenotype Information
- 4.3 Network Weight-based Evolutionary Algorithm
- 4.4 Conclusions
- 5 Experimental Studies
- 5.1 Evolving Connection Weights
- 5.1.1 Experimental Setup
- 5.1.2 Investigating the Impact of a Particular Error in NWEA
- 5.1.3 Convergence Speed: Iterations and Time
- 5.1.4 Percentage of Successful Improvements
- 5.1.5 Increasing Accuracy of the Evolved Solutions
- 5.2 Evolving Connection Weights and Architectures
- 5.2.1 Encoding Scheme for ANN Topologies and Connection Weigths
- 5.2.2 Architecture Mutation During Evolution
- 5.2.3 Data Sets and Experimental Setup
- 5.2.3.1 The Mackey-Glass Chaotic Time Series Problem
- 5.2.3.2 The Breast Cancer Data Set
- 5.2.3.3 The Heart Disease Data Set
- 5.2.3.4 The Diabetes Data Set
- 5.2.3.5 The Thyroid Data Set
- 5.2.4 Evolving ANNs with NWEA: Results and Comparative Analysis
- 5.2.5 Exploring the Impact of Activation Function Type
- 5.2.6 Mixing Different Search Biases in NWEA
- 5.3 Parallelizing NWEA: Investigating Generalization in PEANNs
- 6 Conclusions
- Bibliography