Sie sind auf Seite 1von 1

Homework 1

Due Date: Oct 14, 2013

Problems from Chapter 2 of Textbook


Problems 2, 8, 9, 12, 31, 43

Additional Problem
Suppose a classifier has decision regions R0 and R1 . Is it possible to modify R0 and R1 in such
a way that both PF and PM will reduce?

Computer Problem
Consider a two-class classification problem with P0 = P1 = 1/2. Suppose the class-conditional
probability density function (pdf) p(x|k ) is Gaussian, with N (mk , Ck ), k = 0, 1, with
0
0

0.1 0.0
0.0 0.15

m0 =
and
C0 =

"

"

m1 =

C1 =

"

"

1
0

0.2 0.0
0.0 0.25

(1)
#

(2)

a) Plot function f (x), defined as


f (x) = max{log p(x|k ) + log Pk }.

(3)

b) From the plot in a), can you roughly identify regions R0 , R1 , and the decision boundary?
Plot them.
c) Repeat a) and b) if the covariance matrices are changed to
C0 =

"

0.1 0.0
0.0 0.15

C1 =

"

0.15 0.0
0.0 0.1

d) Repeat a) to c) if P0 is changed to 0.8 (P1 is changed to 0.2).

(4)

Das könnte Ihnen auch gefallen