This Unit introduces machine learning approaches in AI. Machine
learning (esp. its modern subfield, deep learning, which we will cover
in Unit 4) is at the center of modern AI. It is used so often and so
broadly in everybody’s daily life that we barely notice it. In this
Unit, we will provide a gentle introduction to the basic concepts in
machine learning, and then introduce three basic machine learning
algorithms at a very high level. The key point in learning is to
understand conceptually how machine learning works, why it works, when
it works well, and when it does not.
Module Learning Outcomes
After successful completion of this module, you should be able to do
the following (in addition to answering the questions listed below):
Explain the advantages and limitations of machine learning-based AI
(CLO 1)
What are the advantages of machine learning over symbolic AI?
What are the limitations of machine learning compared to symbolic
AI?
What scenarios are particularly suitable for machine learning?
What are overfitting and underfitting? How to prevent them?
Formulate real-world applications as classification or regression
problems (CLO 2)
What is a classification problem?
What is a regression problem?
What is linear classification?
Explain high-level intuitions behind the three basic machine
learning algorithms (nearest neighbor, decision trees, and perceptron)
(CLO 1-2)
How would the \(k\)-nearest
neighbor algorithm be used for handwritten digit classification?
How would you construct a decision tree for a spam filter?
What is the intuition behind the perceptron algorithm?