Weitnauer, Erik: Interactions between perception and rule-construction in human and machine concept learning. 2016
Inhalt
- Abstract
- Publications
- Acknowledgments
- Contents
- 1 Introduction
- 2 Perceiving Physical Scenes
- 2.1 Feature Space
- 2.2 Physical Features
- 2.3 Spatial Relations
- Related Work
- Fuzzy Spatial Relations
- Fuzzy Landscape Algorithm
- Bipolar Fuzzy Landscapes
- Combining Spatial Concepts
- 2.4 Group Attributes
- 2.5 Conclusion
- 3 Learning Physical Concepts
- 3.1 Guiding Principles of Model Design
- 3.2 Implementation
- Scenes
- Switching Between Active Scenes
- Objects and Groups
- Features and Percepts
- Selectors
- Hypotheses
- Attention Mechanisms
- Actions
- 3.3 Utility Estimation
- 3.4 A Problem Solution Walkthrough
- 3.5 Conclusion
- 4 Human Performance on PBPs
- 4.1 The Role of Similarity in Concept Learning
- 4.2 Eye Tracking Study
- 4.3 Scene Ordering Experiments
- Amazon Mechanical Turk as Research Platform
- 1st Experiment
- 2nd Experiment
- 3rd Experiment
- 4th Experiment
- 4.4 General Discussion
- 5 Model Performance on PBPs
- 5.1 Experimental Setup
- 5.2 Results
- Reaction Time Distribution
- Performance Correlation per Problem
- Influence of Presentation Condition
- Efficiency
- 5.3 Discussion
- 6 Conclusion
- A List of Physical Bongard Problems
- Bibliography