Planning of Friday's presentation

A temporary plan for Friday's presentation

Summary of multi-classification methods in both machine learning literature and microarray literature:

Two basic ideas:

1. Classification all at once, based on posterior probability of each class.
Including:

• Parametric: discriminant analysis: LDA, QLDA, Logistic Regression
• Nonparmetric: KNN & Prototype Methods, LVQ, FDA, PDA, Neural Network

2. Binary classification methods + methods for combining binary classification result

• 2-class classification methods: SVM, Golub's Weighted Votes
• Methods of combining binary classification: coding system
  pairwise comparison
  One vs all (OVA)