10-601 Classification and K-NN

This a lecture used in the Syllabus for Machine Learning 10-601 in Fall 2014

Slides

• Ziv's lecture: Slides in pdf.

• William's lecture (draft): Slides in Powerpoint.

Readings

• Mitchell, Chapters 1,2 and 8.

What You Should Know Afterward

• What is the goal of classification
• Bayes decision boundary for classification
• Is there an optimal classifier?
• What the K-NN algorithm is.
• What the computational properties of eager vs lazy learning are in general, and K-NN in specific.
• What decision boundary is defined by K-NN, and how it compares to decision boundaries of linear classifiers.
  • Ziv - shouldn't we move these till after we've introduced an eager learner and a linear classifier? --Wcohen (talk) 13:35, 15 August 2014 (EDT)
• How the value of K affects the tendency of K-NN to overfit or underfit data.
• (optional) probabilistic interpretation of KNN decisions