Feature extraction using fuzzy complete linear discriminant analysis

24
Feature extraction using fuzzy complete linear discriminant analysis The reporter Cui Yan 2012. 4. 26

description

Feature extraction using fuzzy complete linear discriminant analysis. The reporter : Cui Yan. 2012. 4. 26. The report outlines. 1.The fuzzy K-nearest neighbor classifier (FKNN) 2.The fuzzy complete linear discriminant analysis 3.Expriments. The Fuzzy K-nearest neighbor classifier - PowerPoint PPT Presentation

Transcript of Feature extraction using fuzzy complete linear discriminant analysis

Page 1: Feature extraction using fuzzy complete linear discriminant analysis

Feature extraction using fuzzy complete linear discriminant analysis

The reporter : Cui Yan

2012. 4. 26

Page 2: Feature extraction using fuzzy complete linear discriminant analysis

The report outlines

1.The fuzzy K-nearest neighbor classifier (FKNN)

2.The fuzzy complete linear discriminant analysis

3.Expriments

Page 3: Feature extraction using fuzzy complete linear discriminant analysis

The Fuzzy K-nearest neighbor classifier (FKNN)

Page 4: Feature extraction using fuzzy complete linear discriminant analysis

Each sample should be classified

similarly to its surrounding samples,

therefore, a unknown sample could be

predicated by considering the

classification of its nearest neighbor

samples.

The K-nearest neighbor classifier (KNN)

Page 5: Feature extraction using fuzzy complete linear discriminant analysis

KNN tries to classify an unknown sample based

on its k-known classification neighbors.

Page 6: Feature extraction using fuzzy complete linear discriminant analysis

FKNN

Given a sample set , a fuzzy M

-class partition of these vectors specify the membership degrees of each sample corres-ponding to each class.

The membership degree of a training vector to o each of M classes is specified by ,

which is computed by the following steps:ijuijx

},,,{ 21 nxxxX

Page 7: Feature extraction using fuzzy complete linear discriminant analysis

Step 1: Compute the distance matrix between

pairs of feature vectors in the training.

Step 2: Set diagonal elements of this matrix to

infinity (practically place large numeric

values there).

Page 8: Feature extraction using fuzzy complete linear discriminant analysis

Step 3: Sort the distance matrix (treat each of

its column separately) in an ascending

order. Collect the class labels of the

patterns located in the closest neigh-

borhood of the pattern under consi-

deration (as we are concerned with k

neighbors, this returns a list of k

integers).

Page 9: Feature extraction using fuzzy complete linear discriminant analysis

Step 4: Compute the membership grade to class i

for j-th pattern using the expression proposed

in [1].

[1] J.M. Keller, M.R. Gray, J.A. Givens, A fuzzy k-nearest neighbor algorithm, IEEE Trans. Syst.Man Cybernet. 1985, 15(4):580-585

pattern. theof label the if )(*0.49

pattern. theof label the if )(*49.051.0

j-thi

j-thiu

kn

kn

ijij

ij

Page 10: Feature extraction using fuzzy complete linear discriminant analysis

A example for FKNN

No. Feature1 Feature2 class

1 0.2000 0.3000 1

2 0.3000 0.2000 1

3 0.4000 0.3000 1

4 0.5000 0.5000 2

5 0.6000 0.4000 2

6 0.5000 0.6000 2

7 0.7000 0.3000 3

8 0.8000 0.4000 3

9 0.7000 0.5000 3

Page 11: Feature extraction using fuzzy complete linear discriminant analysis

1 2 3 4 5 6 7 8 9

0

0.1414

0.2000

0.3606

0.4123

0.4243

0.5000

0.5385

0.6083

0

0.1414

0.1414

0.3606

0.3606

0.4123

0.4472

0.5000

0.5385

0

0.1414

0.2000

0.2236

0.2236

0.3000

0.3162

0.3606

0.4123

0

0.1000

0.1414

0.2000

0.2236

0.2828

0.3162

0.3606

0.3606

0

0.1414

0.1414

0.1414

0.2000

0.2236

0.2236

0.3606

0.4123

0

0.1000

0.2236

0.2236

0.3162

0.3606

0.3606

0.4243

0.4472

0

0.1414

0.1414

0.2000

0.2828

0.3000

0.3606

0.4123

0.5000

0

0.1414

0.1414

0.2000

0.3162

0.3606

0.4123

0.5385

0.6083

0

0.1414

0.1414

0.2000

0.2000

0.2236

0.3606

0.5000

0.5385

No. 1 2 3 4 5 6 7 8 9

1 0 0.1414 0.2000 0.3606 0.4123 0.4243 0.5000 0.6083 0.5385

2 0.1414 0 0.1414 0.3606 0.3606 0.4472 0.4123 0.5385 0.5000

3 0.2000 0.1414 0 0.2236 0.2236 0.3162 0.3000 0.4123 0.3606

4 0.3606 0.3606 0.2236 0 0.1414 0.1000 0.2828 0.3162 0.2000

5 0.4123 0.3606 0.2236 0.1414 0 0.2236 0.1414 0.2000 0.1414

6 0.4243 0.4472 0.3162 0.1000 0.2236 0 0.3606 0.3606 0.2236

7 0.5000 0.4123 0.3000 0.2828 0.1414 0.3606 0 0.1414 0.2000

8 0.6083 0.5385 0.4123 0.3162 0.2000 0.3606 0.1414 0 0.1414

9 0.5385 0.5000 0.3606 0.2000 0.1414 0.2236 0.2000 0.1414 0

Page 12: Feature extraction using fuzzy complete linear discriminant analysis

1 2 3 4 5 6 7 8 9

1 2 3 4 5 6 7 8 9

2 1 2 6 4 4 5 9 5

3 3 1 5 9 5 8 7 8

4 4 5 9 7 9 9 5 4

5 5 4 3 8 3 4 4 7

6 7 7 7 6 7 3 6 6

7 6 6 8 3

9 9 9 1 2

8

1

6

2

3

2

3

2

8 8 8 2 1 2 1 1 1

1 2 3 4 5 6 7 8 9

1 1 1 2 2 2 3 3 3

1 1 1 2 2 2 2 3 2

1 1 1 2 3 2 3 3 3

2 2 2 3 3 3 3 2 2

2 2 2 1 3 1 2 2 3

2 3 3 3 2 3 1 2 2

3 2 2 3 1

3 3 3 1 1

3

1

2

1

1

1

1

1

3 3 3 1 1 1 1 1 1

Page 13: Feature extraction using fuzzy complete linear discriminant analysis

1 2 3 4 5 6 7 8 9

1 1 1 2 2 2 2 3 2

1 1 1 2 3 2 3 3 3

2 2 2 3 3 3 3 2 2

Set k=3

1 2 3 4 5 6 7 8 9

0.8367 0.8367 0.8367 0 0 0 0 0 0

0.1633 0.1633 0.1633 0.8367 0.6733 0.8367 0.1633 0.1633 0.3267

0 0 0 0.1633 0.3267 0.1633 0.8367 0.8367 0.6733

Page 14: Feature extraction using fuzzy complete linear discriminant analysis

The fuzzy complete linear

discriminant analysis

Page 15: Feature extraction using fuzzy complete linear discriminant analysis

For the training set , we define the i-th class mean by combining the fuzzy membership degree as

And the total mean as

},,,{ 21 nxxxX

.,,2,1 ,1

1 ciu

xum n

j ij

n

j jij

i

1

1

n

i in xm

(1)

(2)

Page 16: Feature extraction using fuzzy complete linear discriminant analysis

Incorporating the fuzzy membership degree, the between-class, the within-class and the total class fuzzy scatter matrix of samples can be defined as

c

i Nj

TjjijtF

c

i Nj

TijijijwF

c

i

n

j

TiiijbF

i

i

mxmxuS

mxmxuS

mmmmuS

1

1

1 1

))((

))((

))((

(3)

Page 17: Feature extraction using fuzzy complete linear discriminant analysis

step1: Calculate the membership degree matrix U

by the FKNN algorithm.

step 2: According toEqs.(1)-(3) work out the

between-class, within-class and total class

fuzzy scatter matrices.

step 3: Work out the orthogonal eigenvectors

p1, . . . , pl of the total class fuzzy scatter

matrix corresponding to positive

eigenvalues.

Algorithm of the fuzzy complete linear analysis

tFS

Page 18: Feature extraction using fuzzy complete linear discriminant analysis

step 4: Let P = (p1, . . . , pl) and

, work out the orthogonal

eigenvectors g1, . . . , gr of correspending

the zero eigenvalues.

step 5: Let P1 = (g1, . . . , gr) and , work

out the orthogonal eigenvectors v1, . . . , vr of

, calculate the irregular discriminant

vectors by .

P, S P S wFT

wF ˆ

P SP S bFT

bF ˆ

1ˆ1~

PS PS bFT

bF

bFS~

vPPwir 1irw

wFS

Page 19: Feature extraction using fuzzy complete linear discriminant analysis

step 6: Work out the orthogonal eigenvectors q1,…, qs

of correspending the non-zero eigenvalues.

step 7: Let P2 = (q1,…, qs) and

, work out the optimal

discriminant vectors vr+1, . . . , vr+s by the

Fisher LDA, calculate the regular discriminant

vectors by .

step 8: (Recognition): Project all samples into the

obtained optimal discriminant vectors and

classify.

wFS

,PS P S wFT

wF 2ˆ2

2ˆ2 PS P S bFT

bF

rw vPPwr 2

Page 20: Feature extraction using fuzzy complete linear discriminant analysis

Experiments

Page 21: Feature extraction using fuzzy complete linear discriminant analysis

• We compare Fuzzy-CLDA with CLDA, UWLDA, FLDA, Fuzzy Fisherface, FIFDA on 3 different data sets from the UCI data sources. The characteristics of the three datasets can be found from (http://archive.ics.uci.edu/ml/datasets).

• All data sets are randomly split to the train set and test set with the ratio 1:4. Experiments are repeated 25 times to obtain mean prediction error rate as a performance measure, NCC is adopted to classify the test samples by using L2 norm.

Page 22: Feature extraction using fuzzy complete linear discriminant analysis
Page 23: Feature extraction using fuzzy complete linear discriminant analysis
Page 24: Feature extraction using fuzzy complete linear discriminant analysis

Thanks!

2012. 4. 26