Visual Media Capture of shape and appearance Preservation of of - - PowerPoint PPT Presentation

Visual Media

Capture of shape and appearance

• f real objects and people

Sign language recognition Preservation of cultural artefacts 3D Broadcast production Animation

Medical Imaging and Remote Sensing

 3D MRI image analysis (brain tumour detection)



Alzheimer’s condition diagnosis (PET brain imaging)



2D-3D Elastic image matching

3D liver reconstruction Microcalcification detection Vascular reconstruction Seismic Pipeline detection

Robot Vision

 Visual learning  Scene interpretation  Model selection  Control of

perception

3D object recognition from 2D views Vision based navigation Target detection Visual surveillance

Multimedia Signal Processing and Interpretation

VOICE FACE LIPS

Fusion

Biometrics Image/Video Retrieval

Ensemble MLP classifier Design

Terry Windeatt University of Surrey, UK

Introduction

 Ensembles – Multiple Classifier Systems (MCS)  Ensemble Multi-layer Perceptron Architecture  Tuning Base Classifiers using measures & OOB estimate  Multi-class ECOC using OOB  Feature Selection and Feature Ranking  Face Recognition

SINGLE CLASSIFIER APPROACH

Goals

• assign a pattern to one of several classes
• find best possible

feature set training set learning machine structure & parameters Task is especially difficult when

• number of classes is high
• classes highly overlapped in feature space
• training samples are few and very noisy

Learning is ill-posed problem & requires built-in assumptions

Multi-layer Perceptron (MLP)

Input layer Hidden layer Output layer Unstable Base Classifier from random starting weights #hidden nodes varies complexity & #epochs varies degree of training

MLP Classifier 1 MLP Classifier 2 MLP Classifier B Combiner

1 2 B

MCS Architecture

Idea is to use multiple simple MLPs rather than single complex MLP Bias/Variance 0/1 loss function more complex than regression

• ensemble reduces variance

& tuning base classifier reduces bias

Multiple Classifiers (MCS)

MCS based upon:

• finding classifiers that perform well but diversely
• appropriate combining strategy

Techniques:

• Different types of classifiers
• Different parameters same classifier
• Different unstable base classifiers e.g MLP
• Different Feature Sets e.g Random Subspace
• Different Training Sets e.g. Bagging/Boosting
• Different class labeling e.g. ECOC

Measures of Diversity

• Accuracy/Diversity Dilemma

Base Classifier Parameter Tuning

 Importance of Parameter Tuning  Every researcher seems to get good results but how?  Need to measure sensitivity to parameters  Helps understand significance of results  Requires systematic change of parameters  How to set parameters?  Alternates to validation set or cross-validation techniques

Out-of-Bootstrap (OOB)

Bootstrapping – Sample with Replacement Promotes diversity among classifiers OOB provides alternative to validation Base classifier OOB uses training patterns left out approx one third Ensemble OOB uses classifiers left out approx one third

 = f(Xm), where m = 1,…. is number of patterns xmi and   {0,1}, i = 1 …B

• 2-class problem
• B parallel base classifiers
• incompletely specified & noisy function

BINARY-TO-BINARY MAPPING

) , , , (

2 1 mB m m m

x x x X  

CLASS SEPARABILITY MEASURE 2-CLASS

calculated over pairs of patterns chosen from different classes

Example: 1 indicates correct classification

0 0 1 1 1 0 0 1 1 0 class 1 1 1 1 0 1 0 0 1 1 0 class 2





 

B j b qj pj a ab pq

N

1

) ( ~  

CLASS SEPARABILITY MEASURE calculated over pairs of patterns p & q chosen from different classes

a,b{0,1}

2 1 00 1 1 11

~ ~ K N K N

q pq q pq

 

 



 



p



, ,

1

y y    

PAIR-WISE DIVERSITY MEASURES Q

b mj m a mi ab ij

N  



  

1

, ,

1

y y    

a,b{0,1}

Use counts between classifier pairs: Giving N11 N10 N01 N00

10 01 00 11 10 01 00 11

N N N N N N N N Q j

i

  

EXPERIMENTS 2-CLASS

 100 single hidden-layer MLP base classifiers  Levenberg-Marquardt training, default parameters  Systematic variation of epochs and nodes  Different random starting weights + bootstrapping  Datasets random 20/80 train/test split (10 runs)

– with added classification noise to encourage overfitting

DATASET #pat #class #con #dis cancer 699 2 9 card 690 2 6 9 credita 690 2 3 11 diabetes 768 2 8 heart 920 2 5 30 ion 351 2 31 3 vote 435 2 16

Figure : Mean test error rates, OOB estimates, measures  , Q for Diabetes 20/80 with [2,4,8,16] nodes

mean test error, , Q over seven 20/80 two-class datasets using 8 hidden- node bootstrapped base classifiers for [0,20,40] % noise

MULTI-CLASS ECOC

 Coding step:

– Map training patterns into two super-classes according to 1’s and 0’s in ECOC matrix Z

 Train base classifier on 2-class decompositions  Decoding step:

– Assign test pattern according to minimum distance to row of ECOC matrix Z

MULTI-CLASS ECOC CODE MATRIX

Example ECOC matrix: 0 1 1 1 ............. 1 0 0 0 ............. 0 1 0 1 ............. 1 0 1 0 ............. 1 1 0 1 ............. 0 0 1 0 ............. each row is a code word each column defines two super-classes

6 classes

Distance-based decoding rules

(e.g. Hamming, L1) 10…1 10…1 1 3 2 Pattern Space ECOC Ensemble Target Classes MLP MLP MLP 01…0 11…1 *** OOB uses only classifiers that are not used in training

Experiments Multi-class

 200 base classifiers  Random ECOC matrices  20/80 train/test split repeated 10 times  Levenberg-Marquardt training algorithm

DATASET #pat #class #con #dis dermatology 366 6 1 33 ecoli 336 8 5 2 glass 214 6 9 iris 150 3 4 segment 2310 7 19 soybean 683 19 35 thyroid 7200 3 6 15 vehicle 846 4 18 vowel 990 11 10 1 wave 5000 3 21 yeast 1484 10 7 1

Yeast 2/4/8/16 nodes 1-69 epochs

Feature Ranking

 Intended for large number of features – small sample  One vs multi-dimensional  Context of MCS - base classifier vs combiner  Simple one-dim methods  Sophisticated multi-dim search methods

Modulus of MLP weights – ‘product of weights’



 

j j ij i

W W w

2 1

W1 is the first layer weight matrix and W2 is the output weight vector

Recursive Feature Elimination

Simple algorithm for eliminating irrelevant features and operates recursively as follows: 1) Rank the features according to a suitable feature ranking method 2) Identify and remove the r least ranked features If r>1, usually desirable from an efficiency viewpoint, a feature subset ranking is obtained.

Mean test error rates, Bias, Variance for RFE MLP ensemble

• ver seven 2-class Datasets 20/80, 10/90. 5/95 train/test split

Yeast RFE for [20/80 10/90 5/95]

Face Recognition - ORL Database

 400 images of forty faces - 40 class identification problem  Variation in lighting,facial hair, pose ….  Controlled background with subjects upright frontal  No need for face detection so fair comparison  We use 40-dim PCA + 20-dim LDA  Random 50/50 train/test split  16 hidden node MLP L-M base classifiers (x200)  Expts repeated twenty times with 40 x 200 ECOC code matrix

ORL Database - Results

Test error, , Q for ORL 50/50 database using 16 hidden-node base classifiers for [0,20,40] % classification noise.

Facial Action Unit (FACS)

 Difficult because depends on age, ethnicity, gender, and

• cclusions due to cosmetics, hair, glasses

 FACS categorises deformation and motion into visual classes  Decouples interpretation from individual actions  Requires skilled practitioners  Small sample size problem

– Large #features and small #training pats

Cohn-Kanade Database

• frontal camera from 100 university students
• contains posed (as opposed to the more difficult

spontaneous) expression sequences

• only the last image is au coded.
• combinations of aus, in some cases non-additive
• Upper face aus au1 au2 au4 au5 au6 au7

Design Decisions

a) All image sequences of size 640 x 480 chosen from the database b) Last image in sequence (no neutral) giving 424 images, 115 containing au1 c) Full image resolution, no compression d) Manually located eye centres plus rotation/scaling into 2 common eye coordinates e) Window extracted of size 150 x 75 pixels centred on eye coordinates f) Forty Gabor filters [18], five special frequencies at five orientations with top 4 principle components for each Gabor filter, 160-dimensional feature vector g) Comparison of feature selection schemes described in Section 3 h) Comparison of MLP ensemble and Support Vector Classifier i) Random training/test split of 90/10 and 50/50 repeated twenty times and averaged

ID sc1 sc2 sc3 sc4 sc5 sc6 superclass {} 1,2 1,2,5 4 6 1,4 #patterns 149 21 44 26 64 18 sc7 sc8 sc9 sc10 sc11 sc12 1,4,7 4,7 4,6,7 6,7 1 1,2,4 10 39 16 7 6 4

ECOC super-classes of action units and number of patterns

2-class Error % 2-class ROC ECOC Error % ECOC ROC au1 8.0/16/28 0.97/16/36 9.0/4/36 0.94/4/17 au2 2.9/1/22 0.99/16/36 3.2/16/22 0.97/1/46 au4 8.5/16/36 0.95//16/28 9.0/1/28 0.95/4/36 au5 5.5/1/46 0.97/1/46 3.5/1/36 0.98/1/36 au6 10.3/4/36 0.94/4/28 12.5/4/28 0.92/1/28 au7 10.3/1/28 0.92/16/60 11.6/4/46 0.92/1/36 mean

7.6 0.96 8.1 0.95

Table 3: Mean best error rates (%) and area under ROC showing #nodes /#features for au classification 90/10 with optimized PCA features and MLP ensemble

Conclusion

 Measures may be used to optimise base classifier

parameters without validation

 OOB estimate can select optimal features

– Even for Ensemble OOB

 Multi-class uses OOB with ECOC  Modulus of MLP weights is simple feature

ranking that works well with RFE

THANK YOU

Feature ranking schemes compared

 RFE with MLP weights  RFE with noisy bootstrap

– Extends training set by resampling with noise

 Boosting single feature each iteration  One-dimensional class-separability

– Trace(SW

• 1 *SB) Within & Between class scatter

 SFFS (Sequential Floating Forward Search)

perceptron-ensemble classifier rfenn rfenb 1dim SFFS boost Mean20/80 15.1 14.6 14.2 15.4 15.4 Mean10/90 16.3 16.3 16.6 18.0 17.6 Mean5/95 18.4 18.5 20.0 21.3 21.3

Table : Mean best error rates for seven two-class problems (20/80, 10/90, 5/95 train/test ) with five feature-ranking schemes

The extended M2VTS (XM2VTS) database

• Contains 295 subjects
• Recorded in four separate sessions over 5 months
• Experimental protocol assigns 200 clients and 95 impostors.
• 3 training, 3 evaluation and 2 test images.
• Impostor set partitioned into 25 evaluation and 70 test impostors
• Features are extracted using PCA + 199-dim LDA

Distance based combination

Use ECOC with 200 x 512 matrix To test client claim is authentic use average distance (L1 Norm) between vector y and the elements of set of class i



 

 

N l b j j j l i

y y N y d

1 1

1 ) (

where yj is the jth binary classifier output, and yl

j is the jth

classifier output for the lth member of class i. distance is checked against a decision threshold

FA 1.3% FR 0.8%

16 node MLP-ensemble classifier rfenn rfenb 1dim SFFS boost 10.0/28 10.9/43 10.9/43 12.3/104 11.9/43 Linear SVC classifier rfesvc rfenb- 1dim SFFS boost 11.6/28 12.1/28 11.9/67 13.9/67 12.4/43

Mean best error rates (%)/number of Gabor features for au1 classification 90/10 with five feature ranking schemes

Windeatt T. and Ghaderi R., Coding and Decoding Strategies for multiclass learning problems, Information Fusion, 4(1), 2003, pp 11-21. Windeatt T, Vote Counting Measures for Ensemble Classifiers, Pattern Recognition, 36(12), 2003, pp 2743- 2756.

• J. Kittler, R. Ghaderi, T. Windeatt and J. Matas Face verification via error correcting output codes, Image and

Vision Computing, Volume 21, Issues 13-14, 1 December 2003, Pages 1163-1169.

• T. Windeatt, Diversity Measures for Multiple Classifier System Analysis and Design, Information Fusion, 6

(1), 2004, 21-36.

• T. Windeatt, Accuracy/ Diversity and Ensemble Classifier Design, IEEE Trans Neural Networks, 17(4), July,

2006.

• R. S. Smith, T. Windeatt, Decoding Rules for ECOC, Proc. 6th Int. Workshop Multiple Classifier Systems,

Editors: N. C. Oza, R. Polikar, J. Kittler, F. Roli, Seaside, Calif, USA, June 2005, Lecture notes in computer science, Springer-Verlag, pp 53-63.

• M. Prior, T. Windeatt, Over-fitting in Ensembles of Neural Network Classifiers within ECOC frameworks,
• Proc. 6th Int. Workshop Multiple Classifier Systems, Editors: N. C. Oza, R. Polikar, J. Kittler, F. Roli,

Seaside, Calif, USA, June 2005, Lecture notes in computer science, Springer-Verlag, pp 286-295.

• T. Windeatt, Ensemble Neural Classifier Design for Face Recognition, European Symposium on Artificial

Neural Networks, ESANN2007, Bruges, April 2007.

• T. Windeatt, M. Prior, Stopping Criteria for Ensemble-based Feature Selection, Proc. 7th Int. Workshop

Multiple Classifier Systems, Prague May 2007, Lecture notes in computer science, Springer-Verlag, pp

• T. Windeatt, M. Prior, N. Effron, N. Intrator, Ensemble-based Feature Selection Criteria, Proc. Conference on

Machine Learning Data Mining MLDM2007, Leipzig, July 2007.