Linear Discriminant Analysis

Classification method

· Based on probabilty of class belonging - models the posterior probability distribution

· Linear decision boundary

Generative Classification via Bayes’ Theorem

Instead of modeling the decision boundary directly, we model the data distribution for each class.

Using Bayes Theorem

$$ P(G=k|X=x)=\frac{f_k(x)π_k}{\sum^K_{l=1}f_l(x)π_1} $$

Where

$$ f_k(x)=P(X=x|G=k) $$

is the class conditional probability.

$$ π_k=P(G=k) $$

is the prior probability of class k.

Basis Expansion