Unit 9: Inductive and Analytic Learning

Commentary

Beginning with this unit, we will explore machine learning - one of the more influential and active fields of AI in recent years. Learning is an essential aspect of advanced intelligent agents or systems, as it is for human beings. Learning includes the ability to improve oneself to gain knowledge and the ability to solve problems, and the ability to adapt to changing environments. Different machine learning principles and techniques have been proposed and applied in the real world, such as inductive learning, analytic learning, statistical learning, and reinforcement learning. In this unit,we explore some learning techniques based on observations and knowledge, such as decision tree learning, ensemble learning, explanation-based learning, relevance-based learning, and inductive logic programming. We will discuss other learning techniques, such as statistical learning and reinforcement learning methods in the units that follow.

Unit Purpose

When you complete this unit, you will be able to

• Define different forms of learning and inductive learning principles.
• Explain the main learning methods, such as decision tree learning, inductive logic programming, and their relevant algorithms.
• Discuss how knowledge and logic are involved in inductive and analytic learning.
• Explain the principles of other learning methods, such as ensemble learning, explanation-based learning, and relevance-based learning.

Section 1: Decision Tree Learning
Section 2: Ensemble learning
Section 3: Logic and knowledge in learning
Section 4: Explanation-based and relevance-based learning
Section 5: Inductive logic programming (ILP)

Readings

Supplemental Unit Readings

Books: Alpaydin, E. (2004). Introduction to machine learning. Cambridge, MA: MIT Press. Mitchell, T. M. (1997). Machine Learning. McGraw-Hill. (ISBN 0-070-42807-7)

Activities

• Explore open source software packages, such as Weka, which include machine learning realizations. Report on the package names, their availability, and capabilities in the course conference.
• Explore logic- and knowledge-based learning strategies, compare their applicability with statistical or reinforcement learning methods, and determine which is currently dominant and which has more potential. Post your ideas and findings to the online course conference.