The key to intelligence is the ability to learn. Studies in the field of machine learning attempt to endow computers with this intrinsic capability that exists in all higher-order organisms to one degree or another. Learning can be loosely defined as a process that improves performance of an agent by acquiring knowledge through interactions with a changing environment.

Areas using machine learning techniques vary from automated knowledge acquisition in expert systems to adaptation in robotic control, from data mining in databases to intelligent information retrieval on the net. Langley (CACM, 11/95) characterizes machine learning into five paradigms: rule learning, analytical learning, case-based learning, genetic algorithms, and neural networks. In this course these paradigms will be explored, and various algorithms and techniques will be discussed. A term project will be required.

Textbook:

• Machine Learning, Tom Mitchell, McGraw-Hill, 1997.

Tentative Schedule:

• Introduction (Ch1)
• Concept Learning (Ch2)
• Decision Tree Learning (Ch3)
• Artificial Neural Networks (Ch4)
• Genetic Algorithms (Ch9)
• Rule Learning (Ch10)
• Instance-Based Learning (Ch8)
• Bayesian Learning (Ch6)

Assignments:

• HW1
• HW2
• HW3
• HW4
• Term paper

Evaluation/Grading:

• Assignments and term project (65%)
• Final exam (35%)

Prerequisites:

• Introduction to AI (CSE 4301 or 5290) or consent of the instructor.
• Basic probability and statistics.
• Proficient knowledge in data structures, algorithms, and programming.

• ACM and IEEE Digital Library
• Citeseer
• Google Scholar