(Copied from old kiwi) |
|||
Line 1: | Line 1: | ||
+ | [http://balthier.ecn.purdue.edu/index.php/ECE662#Course_Topics Course Topics] | ||
+ | |||
One way to represent classifiers. Given classes w1, w2, .., wk and feature vector x, which is in n-dimensional space, the discriminant functions g1(x), g2(x), .., gk(x) where g#(x) maps n-dimensional space to real numbers, are used to make decisions. Decisions are defined as w# if g#(x) >= gj(x) for all j. | One way to represent classifiers. Given classes w1, w2, .., wk and feature vector x, which is in n-dimensional space, the discriminant functions g1(x), g2(x), .., gk(x) where g#(x) maps n-dimensional space to real numbers, are used to make decisions. Decisions are defined as w# if g#(x) >= gj(x) for all j. | ||
Revision as of 17:35, 30 March 2008
One way to represent classifiers. Given classes w1, w2, .., wk and feature vector x, which is in n-dimensional space, the discriminant functions g1(x), g2(x), .., gk(x) where g#(x) maps n-dimensional space to real numbers, are used to make decisions. Decisions are defined as w# if g#(x) >= gj(x) for all j.
Discriminant functions are used to define Decision Surfaces_Old Kiwi.