Elements of Machine LearningRecent years have seen an explosion of work on machine learning, the computational study of algorithms that improve performance based on experience. Research on rule induction, neural networks, genetic algorithms, case-based reasoning, and probabilistic inference has produced a variety of robust methods for inducing knowledge from training data. This book covers the main induction algorithms explored in the literature and presents them within a coherent theoretical framework that moves beyond traditional paradigm boundaries. Elements of Machine Learning provides a comprehensive introduction to the fundamental concepts and problems in the field. The book illustrates a variety of basic algorithms for inducing simple concepts from experience, presents alternatives for organizing learned concepts into large-scale structures, and discusses adaptations of the learning methods to more complex problem-solving tasks. The chapters describe these computational techniques in detail and give examples of their operation, along with exercises and references to the literature. This text is suitable for use in graduate courses on machine learning. Researchers and students in artificial intelligence |
Contents
An overview of machine learning | 1 |
The induction of logical conjunctions | 27 |
The induction of threshold concepts | 67 |
The induction of competitive concepts | 95 |
The construction of decision lists | 115 |
Revision and extension of inference networks | 149 |
The formation of concept hierarchies | 187 |
Other issues in concept induction | 227 |
The formation of transition networks | 257 |
The acquisition of searchcontrol knowledge | 289 |
The formation of macrooperators | 331 |
Prospects for machine learning | 377 |
| 389 | |
| 415 | |
Common terms and phrases
action additional algorithm alternative apply approach associated assumes attributes block boundaries Chapter classifier clear combination component compute concept conjunctions consider construct contains cover creating decision described descriptions determine disjunctive domains effects evaluation examine example existing extension Figure framework function girth given giving goal hypothesis incremental induction inference network initial involves issues knowledge lead literals logical machine learning macro-operators match methods nature negative instances node observed occur operator organization path performance positive positive instances possible prediction present probability problem problem solver produce recursive remaining removing representation requires researchers result revised rules scheme score sequence shows similar simple solution solving space specific stack step stored structure Table task techniques threshold threshold units tion training data training instances tree typically values weights