Real-valued Anticipatory Classifier System
Real-valued Anticipatory Classifier System
1. Introduction
1.1. Motivation and challenges
1.2. Research hypothesis, its aims and goals
1.3. Thesis structure
2. Selected topics of Learning Classifier Systems
2.1. Road towards Anticipatory Learning Classifier Systems
2.2. Real-valued input challenge
2.3. Key Performance Indicators
2.4. Statistical verification of results
2.5. Overview of the selected environments
3. Ways of handling real-valued input signal
3.1. Interval-based representation
Experiment 1 - Encoding precision
Experiment 2 - Nature of the intervals
3.2. Discretizing input signal
Experiment 3 - Single-step environment performance
Experiment 4 - Multiple-step environments performance
4. Optimizing formation of internal model
4.1. Biased exploration
Experiment 1 - Single-step problem performance
Experiment 2 - Multi-steps problems performance
Experiment 3 - Balancing the pole
5. Optimizing reward distribution through long action chains
5.1. Diminishing reward
Experiment 1 - Straight Corridor
Experiment 2 - Deceptive Corridor
6. Summary
6.1. Conclusions
6.2. Future Works
6.3. Publications
7. Abbreviations
8. Bibliography
.md
.pdf
repository
open issue
Optimizing reward distribution through long action chains
5.
Optimizing reward distribution through long action chains
¶
5.1. Diminishing reward
Experiment 1 - Straight Corridor
Experiment 2 - Deceptive Corridor
previous
Experiment 3 - Balancing the pole
next
5.1.
Diminishing reward
