Unformatted text preview: Statistical Pattern Recognition University of Maryland, College Park October 8, 2009 Professor Rama Chellappa Handout 5 Homework 5 This problem set is due Thursday October 15 at 11:00AM . 1. Consider a case in which class ω 1 consists of the two feature vectors [0 , 0] T and [0 , 1] T and class ω 2 of [1 , 0] T and [1 , 1] T . Use the perceptron in it’s reward and punishment form (referred to as the fixed increment single sample perceptron (FISSP) in class), with ρ = 1 (referred to as learning rate η in class) and w (0) = [0 , 0] T , to design the line separating the two classes. 2. Generate 50 feature vectors for each of the two classes p ( x  ω 1 ) ∼ N parenleftbiggparenleftbigg 1 1 parenrightbigg ,σ 2 I parenrightbigg , p ( x  ω 2 ) ∼ N parenleftbiggparenleftbigg parenrightbigg ,σ 2 I parenrightbigg , σ 2 = 0 . 2. To guarantee linear separability of the classes disregard vectors with x 1 + x 2 < 1 for class ω 1 and vectors with x 1 + x 2 > 1 for class ω 2 ....
View
Full
Document
This note was uploaded on 10/26/2009 for the course CMSC 828 taught by Professor Staff during the Fall '05 term at Maryland.
 Fall '05
 staff

Click to edit the document details