10-601B Decision Trees

This a lecture used in the Syllabus for Machine Learning 10-601B in Spring 2016

Readings

• Mitchell, Chapter 3 or Murphy 16.1-16.2

What You Should Know Afterward

• What a decision tree is, and how to classify an instance using a decision tree.
• What the canonical top-down algorithm is for learning a decision tree.
  • What heuristics are used for choosing a decision-tree split.
  • What entropy is, and what information gain is.
• What reduced-error pruning is, and why it might improve classification performance.
• Some of the advantages and disadvantages of decision-tree learning in specific, and eager learning in general, compared to K-NN learning.