Computer Science 456: Artificial Intelligence and Expert Systems

Unit 8: Introduction to Machine Learning

Machine learning usually refers to the changes in systems that perform tasks associated with artificial intelligence. In this context we consider machine learning as the process that allow the machine to change its structure, program, or data based on external information collected using sensors in such a manner that its expected future performance improves. This unit provides an overview of the basic concepts and algorithms for the two main approaches to machine learning: the symbol-based model and the connectionist model.

Activities

Project

Complete the project, and submit it to your tutor for evaluation and feedback.

Section 8.1: The Symbol-based Model for Machine Learning

This section provides an introduction to machine learning in general but focuses on the symbol-based model. Symbol-based learning includes many techniques such as induction, space search, and algorithms such as the ID3. In this section we will focus on inductive learning and present two main algorithms in this area: version space search and ID3.

Learning Objectives

• Discuss a framework for symbol-based learning.
• Present the version space algorithms.
• Present the ID3 decision tree algorithm.

Key Terms

inductive learning, concept space, heuristic search, version space search, candidate elimination algorithm, supervised learning, ID3 decision tree algorithm, information gain

Readings

Read the sections “Introduction,” “A Framework for Symbol-based Learning,” “Version Space Search,” and “The ID3 Decision Tree Induction Algorithm” from Chapter 10, Machine Learning: Symbol-based.

Tasks

Practice the following exercises from Chapter 10 of the textbook:

•  Exercises 2 and 6. You may want to use the Personal Workspace wiki on the course home page and/or share your observations with classmates in the COMP 456 General Discussion forum.

Section 8.2: The Connectionist Model for Machine Learning

This section introduces neural networks that represent the connectionist model of learning. In neural networks the learning evolves from a rearrangement and modification of the overall weighting of nodes and structure of the system. In this section we will introduce the basic concepts of learning using neural networks through the presentation of the perceptron architecture and the backpropagation algorithm.

Learning Objectives

• Discuss the foundational concepts of neural networks.
• Present the perceptron learning algorithm.
• Present the backpropagation algorithm.

Key Terms

artificial neuron, perceptron, delta rule, backpropagation learning

Readings

Read the sections “Introduction,” “Foundations for Connectionist Networks,” “Perceptron Learning,” and “Backpropagation Learning” from Chapter 11, Machine Learning: Connectionist.

Tasks

Practice the following exercises from Chapter 11 of the textbook:

• Exercises 1 and 2. You may want to use the Personal Workspace wiki on the course home page and/or share your observations with classmates in the COMP 456 General Discussion forum.

Final Examination

Arrange to complete the final examination well before the end of your contract or sooner if required for funding agencies, transfer credit, etc.