Use of Extreme Value Statistics in Modeling Biometric Systems

Size: px
Start display at page:

Download "Use of Extreme Value Statistics in Modeling Biometric Systems"

Transcription

1 Use of Extreme Value Statistics in Modeling Biometric Systems

2 Similarity Scores Two types of matching: Genuine sample Imposter sample Matching scores Enrolled sample

3 Probability Density Decision Threshold Imposter sample Imposter 0.32 Genuine Genuine sample 0.95 FRR t 0 t EER FAR matching score Threshold A False Accept (FA) is made when a presented identity is incorrectly accepted (false positive). A False Reject (FR) is a result obtained when the presented identity is incorrectly rejected (false negative). The False Accept Rate (FAR) and False Reject Rate (FRR) together characterize the performance of an authentication system, which can be expressed in terms of the cumulative probability distributions of scores.

4 Objectives Score level modeling in terms of Probability Density Functions (PDFs) or cumulative Distribution Function (CDF) Data: The genuine scores and impostor scores. Methods: Extreme value statistics Issues: Tail distribution estimation Reliability assurance confidence intervals

5 Score level modeling Applications: Multimodal Fusion System performance optimization and evaluation Prediction of how systems scale to very large populations Confidence interval estimation How sure is that the measured performance will hold up. What s the sample variance? How sure is that the measured performance beats a performance requirement? How many samples need to be used for this?

6 Distribution modeling Classical statistics methods: Parametric methods: Gaussian (central limit theorem) mixture models Non-parametric method: Kernel density estimates

7 Distribution modeling Classical statistics methods (continue) Limitation: model the AVERAGE behavior of a process. Rely on the assumption of the nature of the distribution. The analysis is largely insensitive to the tails of the distribution where the biometric systems are prone to error.

8 Distribution modeling Extreme value theory (EVT) Model the EXTREME behavior (the tail of a distribution) Deal with very small data sets!

9 Biometric score modeling Earlier methods: Empirical estimation Given a set of match scores (genuine or imposter) X = {X 1,,X M } drawn from a population with distribution F(x) = Prob(X x). A simple naive estimator called the empirical distribution function is used: where the estimation is simply a count of the X i in X that are smaller than t 0 (divided by M).

10 Biometric score modeling Earlier methods: Empirical estimation F(x) = Prob(X x) (CDF) p(x) (PDF) x

11 Biometric score modeling Earlier methods: Simulation Approach Given a set of match scores Use the same empirical distribution function Define Value of H(t) at is linearly interpolated The estimated distribution of F(t) is then given as

12 Biometric score modeling Earlier methods: Simulation Approach (continue) Let then is actually exponential quantile of the empirical distribution. The estimated distribution is To generate simulation data, one starts with the uniform random variable U in [0, 1]. Data following distribution is then generated as

13 Extreme Value Theory and Tail CDF Why tail distribution important? Most biometric systems expect an equal error rate (EER) of 5% or lower. Many fingerprint verification systems achieve even 2% EER [1]. Threshold setting as well as assessing the claimed performance of vendor systems require accurate estimation of the tail distributions of both genuine and impostor scores (See Fig. 1).

14 Extreme Value Theory and Tail CDF Difficulty: Lack of data in tail. 95% threshold 5% of the scores leaves the tail only 139 scores. 2% tail will leave us with only 56 scores. Empirical distribution of 2800 genuine scores of CUBS data generated on FVC2004 fingerprint images.

15 Extreme Value Theory and Tail CDF EVT is explicitly focused on accurate modeling of the tails of the distributions and thus ideal for modeling biometric matching scores For the biometric scores being truncated at t, leaving the tail data, the distribution function (d.f.) is defined as:

16 Extreme Value Theory (EVT) Approaches Block Maxima (GEV) Threshold approach (GPD)

17 Extreme Value Theory (EVT) Block Maxima (GEV) Example: Biometric system monitoring: Model extreme daily recorded matching scores Take block maximum maximum daily matching scores (minimum of genuine or maximum of imposter) for each day: M n = max{x 1,,X n } m collected daily records (data points for M n, training data):

18 Extreme Value Theory (EVT) Block Maxima (GEV) Extremal Type Theory The distribution of maximum, M n = max{x 1,,X n } converges to (as n ) G(x) is called the Generalized Extreme Value (GEV) distribution and has 3 parameters: shape parameter ξ location parameter µ scale parameter σ.

19 Extreme Value Theory (EVT) Block Maxima (GEV) It is another limit theorem The Central Limit Theorem If the sum of the variables has a finite variance, then it will be approximately normally distributed The extremal type theorem the maximum of many i.i.d random variables has an asymptotic distribution as GEV

20 Extreme Value Theory (EVT) Threshold Models Model exceedances over a high threshold u {Y X>u}, where Y=X-u. Allows to make more efficient use of the data E.g. off daily measurement of biometric matching scores Block maxima use one number Threshold models use all numbers above the threshold

21 Extreme Value Theory (EVT) Threshold Models Specifically for modeling tail distribution For accurate estimation of the tail

22 Extreme Value Theory (EVT) Threshold Models for tail estimation Let X 1,X 2, be a sequence of i.i.d. random variables having a distribution function F Consider the sequence of sample excess Y 1, Y 2 where Y i = X i - u is defined for all X i > u The conditional excess distribution function is then defined as

23 Extreme Value Theory (EVT) Threshold Models Theorem: The distribution of Y :=X-u X>u converges to (as u ) is called the Generalized Pareto distribution (GPD) with 2 parameters. shape parameter ξ scale parameter σ. The shape parameter ξ is the same parameter as in the GEV distribution.

24 Extreme Value Theory (EVT) Threshold Models theory

25 Extreme Value Theory (EVT) Tail estimation of the underlying distribution The relation between conditional excess d.f. F u and the underlying distribution F Therefore for x u (the tail), (remember y=x-u)

26 Extreme Value Theory (EVT) Tail estimated by GPD: With the empirical distribution value at u) and for F(u) (CDF evaluated at one where N u is the number of data above the threshold u. We have Which simplifies to

27 Extreme Value Theory (EVT) Tail estimated by GPD: In two-parameter case where Nu is the number of data above the threshold u

28 Fitting biometric data to GPD Fitting procedure Data collection Exploratory graphical tools Quantile-Quantile Plot (QQ-Plot) Sample Mean Excess Plot Fitting: Maximum Likelihood Estimation

29 Fitting biometric data to GPD Data collection CUBS Fingerprint Matching Data 2800 genuine matching scores and 4950 impostor matching scores NIST biometric matching scores Fingerprint, iris, face Ultra-scan Fingerprint Matching Data Training and testing Both contain impostor and 1000 genuine scores

30 Fitting biometric data to GPD Data collection Histogram of the CUBS genuine scores (flipped) Sample size is 2800

31 Fitting biometric data to GPD Exploratory graphical tools Quantile-Quantile Plot (QQ-plot) Validating that if a data set comes from populations of a given distribution. A quantile: Inverse of CDF A reference line is also plotted. If the two sets come from a population with the same distribution, the points should fall approximately along this reference line.

32 Fitting biometric data to GPD Quantile-Quantile Plot (QQ-plot) The QQ-plot against the exponential distribution is a very useful guide for visual examination of the hypothesis that the data comes from an exponential distribution, i.e. a special case of GPD. QQ-plot of the quantiles of the empirical distribution function against exponential distribution is defined as Where X k,n is k-th value in the ordered list and G is the exponential distribution function. (the considered data set is ordered as {X 1:n,, X n:n } )

33 Fitting biometric data to GPD Quantile-Quantile Plot (QQ-plot) QQ-plot of the CUBS data plotted against the exponential distribution

34 Fitting biometric data to GPD Sample Mean Excess Plot For threshold selection Theorem: If X has a GPD with parameters ξ and σ, the excess over the threshold u also has a GPD and its mean excess function over this threshold is given by [13] which is a linear function of u. If the empirical mean excess of given scores has a linear shape, then a GPD might be the model for the overall distribution

35 Fitting biometric data to GPD Sample Mean Excess Plot In the case when the mean excess plot has a linear shape only beyond a certain point, it suggests that the point should be a candidate for a threshold. The sample mean plot is defined as the plot of the points: Where the data set is sorted as {X 1:n,, X n:n } in ascendant order and the sample mean excess function is defined as

36 Fitting biometric data to GPD Sample Mean Excess Plot The sample mean excess plot of CUBs data shows a general linear pattern. A close inspection of the plot suggests that a threshold can be chosen at the value u = 0.45

37 Fitting biometric data to GPD Fitting Maximum Likelihood Estimation Taking first derivative of GPD for a sample x = {x 1,, x n }, the log-likelihood function for the GPD n L(, x) log G, ( x i ) Therefore i 1 a

38 Confidence interval for CDF using GPD Derive Close form confidence interval for the tail estimation using GPD Recall the tail estimation Rename it as

39 Confidence interval for CDF using GPD The Taylor expansion around the estimated The variance of is given asymptotically

40 Confidence interval for CDF using GPD The Fisher information matrix gives the asymptotic variance of the MLE (Smith 1984 [18]) Thus For CEF with asymptotic confidence level α, the (1- α)100% confidence interval is

41 Experiments Fitting with GPD GPD fits to imposter data of the training set values GPD fit to tails of imposter data The red lines are the GPD estimates, which are inside the empirical estimates of the data represented by solid darker dots. The thick vertical bars are examples of confidence intervals.

42 Experiments Fitting with GPD GPD fits to genuine data of the training set values GPD fit to tails of genuine data The red lines are the GPD estimates, which are inside the empirical estimates of the data represented by solid darker dots. The thick vertical bars are examples of confidence intervals.

43 Experiments Fitting with GPD Test imposter fitting: The test CDFs (solid darker dots) are matching or close to the estimates of training CDFs (green line). The thick vertical bar again is an example of confidence intervals. (The confidence for imposter data is too small to see in the plot.)

44 Experiments Fitting with GPD Test imposter tail fitting:

45 Experiments Fitting with GPD Test genuine fitting: The test CDFs (solid darker dots) are matching or close to the estimates of training CDFs (green line). The thick vertical bar again is an example of confidence intervals.

46 Experiments Fitting with GPD ROC: ROC with an example of confidence box. Note that the box is narrow in the direction of imposter score due to ample availability of data.

47 Conclusion EVT is the most scientific approach to an inherently difficult problem predicting the probability of a rare event. Extreme Value Theory is a theoretically supported method for problems with interest in accurately modeling the tails of the probability distributions. EVT is a feasible method being used for biometric data. The derivation of the confidence interval for GPD estimation and confidence bands for ROC, is important contribution to the field of biometrics.

Extreme Value Theory in (Hourly) Precipitation

Extreme Value Theory in (Hourly) Precipitation Extreme Value Theory in (Hourly) Precipitation Uli Schneider Geophysical Statistics Project, NCAR GSP Miniseries at CSU November 17, 2003 Outline Project overview Extreme value theory 101 Applying extreme

More information

Chapter 3. Bootstrap. 3.1 Introduction. 3.2 The general idea

Chapter 3. Bootstrap. 3.1 Introduction. 3.2 The general idea Chapter 3 Bootstrap 3.1 Introduction The estimation of parameters in probability distributions is a basic problem in statistics that one tends to encounter already during the very first course on the subject.

More information

Client Dependent GMM-SVM Models for Speaker Verification

Client Dependent GMM-SVM Models for Speaker Verification Client Dependent GMM-SVM Models for Speaker Verification Quan Le, Samy Bengio IDIAP, P.O. Box 592, CH-1920 Martigny, Switzerland {quan,bengio}@idiap.ch Abstract. Generative Gaussian Mixture Models (GMMs)

More information

Probability Models.S4 Simulating Random Variables

Probability Models.S4 Simulating Random Variables Operations Research Models and Methods Paul A. Jensen and Jonathan F. Bard Probability Models.S4 Simulating Random Variables In the fashion of the last several sections, we will often create probability

More information

Continuous Improvement Toolkit. Normal Distribution. Continuous Improvement Toolkit.

Continuous Improvement Toolkit. Normal Distribution. Continuous Improvement Toolkit. Continuous Improvement Toolkit Normal Distribution The Continuous Improvement Map Managing Risk FMEA Understanding Performance** Check Sheets Data Collection PDPC RAID Log* Risk Analysis* Benchmarking***

More information

A Framework for Efficient Fingerprint Identification using a Minutiae Tree

A Framework for Efficient Fingerprint Identification using a Minutiae Tree A Framework for Efficient Fingerprint Identification using a Minutiae Tree Praveer Mansukhani February 22, 2008 Problem Statement Developing a real-time scalable minutiae-based indexing system using a

More information

Off-Line and Real-Time Methods for ML-PDA Target Validation

Off-Line and Real-Time Methods for ML-PDA Target Validation Off-Line and Real-Time Methods for ML-PDA Target Validation Wayne R. Blanding*, Member, IEEE, Peter K. Willett, Fellow, IEEE and Yaakov Bar-Shalom, Fellow, IEEE 1 Abstract We present two procedures for

More information

Cpk: What is its Capability? By: Rick Haynes, Master Black Belt Smarter Solutions, Inc.

Cpk: What is its Capability? By: Rick Haynes, Master Black Belt Smarter Solutions, Inc. C: What is its Capability? By: Rick Haynes, Master Black Belt Smarter Solutions, Inc. C is one of many capability metrics that are available. When capability metrics are used, organizations typically provide

More information

DOT-K: Distributed Online Top-K Elements Algorithm with Extreme Value Statistics

DOT-K: Distributed Online Top-K Elements Algorithm with Extreme Value Statistics DOT-K: Distributed Online Top-K Elements Algorithm with Extreme Value Statistics Nick Carey, Tamás Budavári, Yanif Ahmad, Alexander Szalay Johns Hopkins University Department of Computer Science ncarey4@jhu.edu

More information

The Expected Performance Curve: a New Assessment Measure for Person Authentication

The Expected Performance Curve: a New Assessment Measure for Person Authentication The Expected Performance Curve: a New Assessment Measure for Person Authentication Samy Bengio Johnny Mariéthoz IDIAP CP 592, rue du Simplon4 192 Martigny, Switzerland {bengio,marietho}@idiap.ch Abstract

More information

Natural Hazards and Extreme Statistics Training R Exercises

Natural Hazards and Extreme Statistics Training R Exercises Natural Hazards and Extreme Statistics Training R Exercises Hugo Winter Introduction The aim of these exercises is to aide understanding of the theory introduced in the lectures by giving delegates a chance

More information

CHAPTER 6 RESULTS AND DISCUSSIONS

CHAPTER 6 RESULTS AND DISCUSSIONS 151 CHAPTER 6 RESULTS AND DISCUSSIONS In this chapter the performance of the personal identification system on the PolyU database is presented. The database for both Palmprint and Finger Knuckle Print

More information

INTERPRETING FINGERPRINT AUTHENTICATION PERFORMANCE TECHNICAL WHITE PAPER

INTERPRETING FINGERPRINT AUTHENTICATION PERFORMANCE TECHNICAL WHITE PAPER INTERPRETING FINGERPRINT AUTHENTICATION PERFORMANCE TECHNICAL WHITE PAPER Fidelica Microsystems, Inc. 423 Dixon Landing Rd. Milpitas, CA 95035 1 INTRODUCTION The fingerprint-imaging segment of the biometrics

More information

Instance-based Learning CE-717: Machine Learning Sharif University of Technology. M. Soleymani Fall 2015

Instance-based Learning CE-717: Machine Learning Sharif University of Technology. M. Soleymani Fall 2015 Instance-based Learning CE-717: Machine Learning Sharif University of Technology M. Soleymani Fall 2015 Outline Non-parametric approach Unsupervised: Non-parametric density estimation Parzen Windows K-Nearest

More information

Today. Lecture 4: Last time. The EM algorithm. We examine clustering in a little more detail; we went over it a somewhat quickly last time

Today. Lecture 4: Last time. The EM algorithm. We examine clustering in a little more detail; we went over it a somewhat quickly last time Today Lecture 4: We examine clustering in a little more detail; we went over it a somewhat quickly last time The CAD data will return and give us an opportunity to work with curves (!) We then examine

More information

Beyond parametric score normalisation in biometric verification systems

Beyond parametric score normalisation in biometric verification systems Published in IET Biometrics Received on 6th October 2013 Revised on 20th February 2014 Accepted on 4th March 2014 Special Issue: Integration of Biometrics and Forensics ISSN 2047-4938 Beyond parametric

More information

Exploring Similarity Measures for Biometric Databases

Exploring Similarity Measures for Biometric Databases Exploring Similarity Measures for Biometric Databases Praveer Mansukhani, Venu Govindaraju Center for Unified Biometrics and Sensors (CUBS) University at Buffalo {pdm5, govind}@buffalo.edu Abstract. Currently

More information

CREATING THE DISTRIBUTION ANALYSIS

CREATING THE DISTRIBUTION ANALYSIS Chapter 12 Examining Distributions Chapter Table of Contents CREATING THE DISTRIBUTION ANALYSIS...176 BoxPlot...178 Histogram...180 Moments and Quantiles Tables...... 183 ADDING DENSITY ESTIMATES...184

More information

DEFORMABLE MATCHING OF HAND SHAPES FOR USER VERIFICATION. Ani1 K. Jain and Nicolae Duta

DEFORMABLE MATCHING OF HAND SHAPES FOR USER VERIFICATION. Ani1 K. Jain and Nicolae Duta DEFORMABLE MATCHING OF HAND SHAPES FOR USER VERIFICATION Ani1 K. Jain and Nicolae Duta Department of Computer Science and Engineering Michigan State University, East Lansing, MI 48824-1026, USA E-mail:

More information

Bootstrapping Method for 14 June 2016 R. Russell Rhinehart. Bootstrapping

Bootstrapping Method for  14 June 2016 R. Russell Rhinehart. Bootstrapping Bootstrapping Method for www.r3eda.com 14 June 2016 R. Russell Rhinehart Bootstrapping This is extracted from the book, Nonlinear Regression Modeling for Engineering Applications: Modeling, Model Validation,

More information

Statistics I 2011/2012 Notes about the third Computer Class: Simulation of samples and goodness of fit; Central Limit Theorem; Confidence intervals.

Statistics I 2011/2012 Notes about the third Computer Class: Simulation of samples and goodness of fit; Central Limit Theorem; Confidence intervals. Statistics I 2011/2012 Notes about the third Computer Class: Simulation of samples and goodness of fit; Central Limit Theorem; Confidence intervals. In this Computer Class we are going to use Statgraphics

More information

Biometric quality for error suppression

Biometric quality for error suppression Biometric quality for error suppression Elham Tabassi NIST 22 July 2010 1 outline - Why measure quality? - What is meant by quality? - What are they good for? - What are the challenges in quality computation?

More information

Comparison of ROC-based and likelihood methods for fingerprint verification

Comparison of ROC-based and likelihood methods for fingerprint verification Comparison of ROC-based and likelihood methods for fingerprint verification Sargur Srihari, Harish Srinivasan, Matthew Beal, Prasad Phatak and Gang Fang Department of Computer Science and Engineering University

More information

The Expected Performance Curve: a New Assessment Measure for Person Authentication

The Expected Performance Curve: a New Assessment Measure for Person Authentication R E S E A R C H R E P O R T I D I A P The Expected Performance Curve: a New Assessment Measure for Person Authentication Samy Bengio 1 Johnny Mariéthoz 2 IDIAP RR 03-84 March 10, 2004 submitted for publication

More information

Introduction to Mobile Robotics Bayes Filter Particle Filter and Monte Carlo Localization. Wolfram Burgard

Introduction to Mobile Robotics Bayes Filter Particle Filter and Monte Carlo Localization. Wolfram Burgard Introduction to Mobile Robotics Bayes Filter Particle Filter and Monte Carlo Localization Wolfram Burgard 1 Motivation Recall: Discrete filter Discretize the continuous state space High memory complexity

More information

Univariate Extreme Value Analysis. 1 Block Maxima. Practice problems using the extremes ( 2.0 5) package. 1. Pearson Type III distribution

Univariate Extreme Value Analysis. 1 Block Maxima. Practice problems using the extremes ( 2.0 5) package. 1. Pearson Type III distribution Univariate Extreme Value Analysis Practice problems using the extremes ( 2.0 5) package. 1 Block Maxima 1. Pearson Type III distribution (a) Simulate 100 maxima from samples of size 1000 from the gamma

More information

Markov chain Monte Carlo methods

Markov chain Monte Carlo methods Markov chain Monte Carlo methods (supplementary material) see also the applet http://www.lbreyer.com/classic.html February 9 6 Independent Hastings Metropolis Sampler Outline Independent Hastings Metropolis

More information

Biometrics Technology: Multi-modal (Part 2)

Biometrics Technology: Multi-modal (Part 2) Biometrics Technology: Multi-modal (Part 2) References: At the Level: [M7] U. Dieckmann, P. Plankensteiner and T. Wagner, "SESAM: A biometric person identification system using sensor fusion ", Pattern

More information

International Journal of Scientific & Engineering Research, Volume 7, Issue 11, November ISSN

International Journal of Scientific & Engineering Research, Volume 7, Issue 11, November ISSN International Journal of Scientific & Engineering Research, Volume 7, Issue 11, November-2016 483 On Securing Automatic Teller Machine Transactions Using Bank Verification Number Ojulari Hakeem, Oke Alice

More information

Application of the Peaks-Over-Threshold Method on Insurance Data

Application of the Peaks-Over-Threshold Method on Insurance Data U.U.D.M. Project Report 2018:32 Application of the Peaks-Over-Threshold Method on Insurance Data Max Rydman Examensarbete i matematik, 15 hp Handledare: Jesper Rydén Examinator: Veronica Crispin Quinonez

More information

Data Mining Classification: Bayesian Decision Theory

Data Mining Classification: Bayesian Decision Theory Data Mining Classification: Bayesian Decision Theory Lecture Notes for Chapter 2 R. O. Duda, P. E. Hart, and D. G. Stork, Pattern classification, 2nd ed. New York: Wiley, 2001. Lecture Notes for Chapter

More information

Combining Biometric Scores in Identification Systems

Combining Biometric Scores in Identification Systems 1 Combining Biometric Scores in Identification Systems Sergey Tulyakov and Venu Govindaraju, Fellow, IEEE Both authors are with the State University of New York at Buffalo. 2 Abstract Combination approaches

More information

MULTI-FINGER PENETRATION RATE AND ROC VARIABILITY FOR AUTOMATIC FINGERPRINT IDENTIFICATION SYSTEMS

MULTI-FINGER PENETRATION RATE AND ROC VARIABILITY FOR AUTOMATIC FINGERPRINT IDENTIFICATION SYSTEMS MULTI-FINGER PENETRATION RATE AND ROC VARIABILITY FOR AUTOMATIC FINGERPRINT IDENTIFICATION SYSTEMS I. Introduction James L. Wayman, Director U.S. National Biometric Test Center College of Engineering San

More information

Statistical Performance Comparisons of Computers

Statistical Performance Comparisons of Computers Tianshi Chen 1, Yunji Chen 1, Qi Guo 1, Olivier Temam 2, Yue Wu 1, Weiwu Hu 1 1 State Key Laboratory of Computer Architecture, Institute of Computing Technology (ICT), Chinese Academy of Sciences, Beijing,

More information

INTENSITY TRANSFORMATION AND SPATIAL FILTERING

INTENSITY TRANSFORMATION AND SPATIAL FILTERING 1 INTENSITY TRANSFORMATION AND SPATIAL FILTERING Lecture 3 Image Domains 2 Spatial domain Refers to the image plane itself Image processing methods are based and directly applied to image pixels Transform

More information

Laplace Transform of a Lognormal Random Variable

Laplace Transform of a Lognormal Random Variable Approximations of the Laplace Transform of a Lognormal Random Variable Joint work with Søren Asmussen & Jens Ledet Jensen The University of Queensland School of Mathematics and Physics August 1, 2011 Conference

More information

Mixture Models and the EM Algorithm

Mixture Models and the EM Algorithm Mixture Models and the EM Algorithm Padhraic Smyth, Department of Computer Science University of California, Irvine c 2017 1 Finite Mixture Models Say we have a data set D = {x 1,..., x N } where x i is

More information

Section 4 Matching Estimator

Section 4 Matching Estimator Section 4 Matching Estimator Matching Estimators Key Idea: The matching method compares the outcomes of program participants with those of matched nonparticipants, where matches are chosen on the basis

More information

Monte Carlo Methods and Statistical Computing: My Personal E

Monte Carlo Methods and Statistical Computing: My Personal E Monte Carlo Methods and Statistical Computing: My Personal Experience Department of Mathematics & Statistics Indian Institute of Technology Kanpur November 29, 2014 Outline Preface 1 Preface 2 3 4 5 6

More information

Slice sampler algorithm for generalized Pareto distribution

Slice sampler algorithm for generalized Pareto distribution Slice sampler algorithm for generalized Pareto distribution Mohammad Rostami, Mohd Bakri Adam Yahya, Mohamed Hisham Yahya, Noor Akma Ibrahim Abstract In this paper, we developed the slice sampler algorithm

More information

Outline. Incorporating Biometric Quality In Multi-Biometrics FUSION. Results. Motivation. Image Quality: The FVC Experience

Outline. Incorporating Biometric Quality In Multi-Biometrics FUSION. Results. Motivation. Image Quality: The FVC Experience Incorporating Biometric Quality In Multi-Biometrics FUSION QUALITY Julian Fierrez-Aguilar, Javier Ortega-Garcia Biometrics Research Lab. - ATVS Universidad Autónoma de Madrid, SPAIN Loris Nanni, Raffaele

More information

Online Signature Verification Technique

Online Signature Verification Technique Volume 3, Issue 1 ISSN: 2320-5288 International Journal of Engineering Technology & Management Research Journal homepage: www.ijetmr.org Online Signature Verification Technique Ankit Soni M Tech Student,

More information

Bootstrapping Methods

Bootstrapping Methods Bootstrapping Methods example of a Monte Carlo method these are one Monte Carlo statistical method some Bayesian statistical methods are Monte Carlo we can also simulate models using Monte Carlo methods

More information

LAB #2: SAMPLING, SAMPLING DISTRIBUTIONS, AND THE CLT

LAB #2: SAMPLING, SAMPLING DISTRIBUTIONS, AND THE CLT NAVAL POSTGRADUATE SCHOOL LAB #2: SAMPLING, SAMPLING DISTRIBUTIONS, AND THE CLT Statistics (OA3102) Lab #2: Sampling, Sampling Distributions, and the Central Limit Theorem Goal: Use R to demonstrate sampling

More information

Regression III: Advanced Methods

Regression III: Advanced Methods Lecture 3: Distributions Regression III: Advanced Methods William G. Jacoby Michigan State University Goals of the lecture Examine data in graphical form Graphs for looking at univariate distributions

More information

Descriptive Statistics, Standard Deviation and Standard Error

Descriptive Statistics, Standard Deviation and Standard Error AP Biology Calculations: Descriptive Statistics, Standard Deviation and Standard Error SBI4UP The Scientific Method & Experimental Design Scientific method is used to explore observations and answer questions.

More information

Improving Personal Identification Accuracy Using Multisensor Fusion for Building Access Control Applications

Improving Personal Identification Accuracy Using Multisensor Fusion for Building Access Control Applications Improving Personal Identification Accuracy Using Multisensor Fusion for Building Access Control Applications Lisa Osadciw, Pramod Varshney, and Kalyan Veeramachaneni laosadci,varshney,kveerama@syr.edu

More information

Week 7: The normal distribution and sample means

Week 7: The normal distribution and sample means Week 7: The normal distribution and sample means Goals Visualize properties of the normal distribution. Learning the Tools Understand the Central Limit Theorem. Calculate sampling properties of sample

More information

Optimal designs for comparing curves

Optimal designs for comparing curves Optimal designs for comparing curves Holger Dette, Ruhr-Universität Bochum Maria Konstantinou, Ruhr-Universität Bochum Kirsten Schorning, Ruhr-Universität Bochum FP7 HEALTH 2013-602552 Outline 1 Motivation

More information

EE795: Computer Vision and Intelligent Systems

EE795: Computer Vision and Intelligent Systems EE795: Computer Vision and Intelligent Systems Spring 2012 TTh 17:30-18:45 WRI C225 Lecture 04 130131 http://www.ee.unlv.edu/~b1morris/ecg795/ 2 Outline Review Histogram Equalization Image Filtering Linear

More information

Lecture 11: Distributions as Models October 2014

Lecture 11: Distributions as Models October 2014 Lecture 11: Distributions as Models 36-350 1 October 2014 Previously R functions for regression models R functions for probability distributions Agenda Distributions from data Review of R for theoretical

More information

Level-set MCMC Curve Sampling and Geometric Conditional Simulation

Level-set MCMC Curve Sampling and Geometric Conditional Simulation Level-set MCMC Curve Sampling and Geometric Conditional Simulation Ayres Fan John W. Fisher III Alan S. Willsky February 16, 2007 Outline 1. Overview 2. Curve evolution 3. Markov chain Monte Carlo 4. Curve

More information

Performance Generalization in Biometric. Sample Bootstraps

Performance Generalization in Biometric. Sample Bootstraps Performance Generalization in Biometric 1 Authentication Using Joint User-Specific and Sample Bootstraps Norman Poh, Alvin Martin and Samy Bengio Abstract Biometric authentication performance is often

More information

Kernel Density Estimation (KDE)

Kernel Density Estimation (KDE) Kernel Density Estimation (KDE) Previously, we ve seen how to use the histogram method to infer the probability density function (PDF) of a random variable (population) using a finite data sample. In this

More information

Frequency Distributions

Frequency Distributions Displaying Data Frequency Distributions After collecting data, the first task for a researcher is to organize and summarize the data so that it is possible to get a general overview of the results. Remember,

More information

BIOMET: A Multimodal Biometric Authentication System for Person Identification and Verification using Fingerprint and Face Recognition

BIOMET: A Multimodal Biometric Authentication System for Person Identification and Verification using Fingerprint and Face Recognition BIOMET: A Multimodal Biometric Authentication System for Person Identification and Verification using Fingerprint and Face Recognition Hiren D. Joshi Phd, Dept. of Computer Science Rollwala Computer Centre

More information

CHAPTER 6. The Normal Probability Distribution

CHAPTER 6. The Normal Probability Distribution The Normal Probability Distribution CHAPTER 6 The normal probability distribution is the most widely used distribution in statistics as many statistical procedures are built around it. The central limit

More information

Keywords Wavelet decomposition, SIFT, Unibiometrics, Multibiometrics, Histogram Equalization.

Keywords Wavelet decomposition, SIFT, Unibiometrics, Multibiometrics, Histogram Equalization. Volume 3, Issue 7, July 2013 ISSN: 2277 128X International Journal of Advanced Research in Computer Science and Software Engineering Research Paper Available online at: www.ijarcsse.com Secure and Reliable

More information

Note Set 4: Finite Mixture Models and the EM Algorithm

Note Set 4: Finite Mixture Models and the EM Algorithm Note Set 4: Finite Mixture Models and the EM Algorithm Padhraic Smyth, Department of Computer Science University of California, Irvine Finite Mixture Models A finite mixture model with K components, for

More information

Part I. Hierarchical clustering. Hierarchical Clustering. Hierarchical clustering. Produces a set of nested clusters organized as a

Part I. Hierarchical clustering. Hierarchical Clustering. Hierarchical clustering. Produces a set of nested clusters organized as a Week 9 Based in part on slides from textbook, slides of Susan Holmes Part I December 2, 2012 Hierarchical Clustering 1 / 1 Produces a set of nested clusters organized as a Hierarchical hierarchical clustering

More information

Biometric Security Roles & Resources

Biometric Security Roles & Resources Biometric Security Roles & Resources Part 1 Biometric Systems Skip Linehan Biometrics Systems Architect, Raytheon Intelligence and Information Systems Outline Biometrics Overview Biometric Architectures

More information

Deficiencies in NIST Fingerprint Image Quality Algorithm Predicting biometric performance using image quality metrics

Deficiencies in NIST Fingerprint Image Quality Algorithm Predicting biometric performance using image quality metrics Deficiencies in NIST Fingerprint Image Quality Algorithm Predicting biometric performance using image quality metrics Martin Aastrup Olsen, Hochschule Darmstadt, CASED 12. Deutscher IT-Sicherheitskongress,

More information

Large Scale Data Analysis Using Deep Learning

Large Scale Data Analysis Using Deep Learning Large Scale Data Analysis Using Deep Learning Machine Learning Basics - 1 U Kang Seoul National University U Kang 1 In This Lecture Overview of Machine Learning Capacity, overfitting, and underfitting

More information

Non-Parametric Modeling

Non-Parametric Modeling Non-Parametric Modeling CE-725: Statistical Pattern Recognition Sharif University of Technology Spring 2013 Soleymani Outline Introduction Non-Parametric Density Estimation Parzen Windows Kn-Nearest Neighbor

More information

Section 6.2: Generating Discrete Random Variates

Section 6.2: Generating Discrete Random Variates Section 6.2: Generating Discrete Random Variates Discrete-Event Simulation: A First Course c 2006 Pearson Ed., Inc. 0-13-142917-5 Discrete-Event Simulation: A First Course Section 6.2: Generating Discrete

More information

CPSC 340: Machine Learning and Data Mining. Non-Parametric Models Fall 2016

CPSC 340: Machine Learning and Data Mining. Non-Parametric Models Fall 2016 CPSC 340: Machine Learning and Data Mining Non-Parametric Models Fall 2016 Admin Course add/drop deadline tomorrow. Assignment 1 is due Friday. Setup your CS undergrad account ASAP to use Handin: https://www.cs.ubc.ca/getacct

More information

CPSC 340: Machine Learning and Data Mining. Non-Parametric Models Fall 2016

CPSC 340: Machine Learning and Data Mining. Non-Parametric Models Fall 2016 CPSC 340: Machine Learning and Data Mining Non-Parametric Models Fall 2016 Assignment 0: Admin 1 late day to hand it in tonight, 2 late days for Wednesday. Assignment 1 is out: Due Friday of next week.

More information

Lab 5 - Risk Analysis, Robustness, and Power

Lab 5 - Risk Analysis, Robustness, and Power Type equation here.biology 458 Biometry Lab 5 - Risk Analysis, Robustness, and Power I. Risk Analysis The process of statistical hypothesis testing involves estimating the probability of making errors

More information

Modelling of non-gaussian tails of multiple Coulomb scattering in track fitting with a Gaussian-sum filter

Modelling of non-gaussian tails of multiple Coulomb scattering in track fitting with a Gaussian-sum filter Modelling of non-gaussian tails of multiple Coulomb scattering in track fitting with a Gaussian-sum filter A. Strandlie and J. Wroldsen Gjøvik University College, Norway Outline Introduction A Gaussian-sum

More information

a. divided by the. 1) Always round!! a) Even if class width comes out to a, go up one.

a. divided by the. 1) Always round!! a) Even if class width comes out to a, go up one. Probability and Statistics Chapter 2 Notes I Section 2-1 A Steps to Constructing Frequency Distributions 1 Determine number of (may be given to you) a Should be between and classes 2 Find the Range a The

More information

An Introduction to the Bootstrap

An Introduction to the Bootstrap An Introduction to the Bootstrap Bradley Efron Department of Statistics Stanford University and Robert J. Tibshirani Department of Preventative Medicine and Biostatistics and Department of Statistics,

More information

Remote Sensing & Photogrammetry W4. Beata Hejmanowska Building C4, room 212, phone:

Remote Sensing & Photogrammetry W4. Beata Hejmanowska Building C4, room 212, phone: Remote Sensing & Photogrammetry W4 Beata Hejmanowska Building C4, room 212, phone: +4812 617 22 72 605 061 510 galia@agh.edu.pl 1 General procedures in image classification Conventional multispectral classification

More information

Introduction to Pattern Recognition Part II. Selim Aksoy Bilkent University Department of Computer Engineering

Introduction to Pattern Recognition Part II. Selim Aksoy Bilkent University Department of Computer Engineering Introduction to Pattern Recognition Part II Selim Aksoy Bilkent University Department of Computer Engineering saksoy@cs.bilkent.edu.tr RETINA Pattern Recognition Tutorial, Summer 2005 Overview Statistical

More information

Peg-Free Hand Geometry Verification System

Peg-Free Hand Geometry Verification System Peg-Free Hand Geometry Verification System Pavan K Rudravaram Venu Govindaraju Center for Unified Biometrics and Sensors (CUBS), University at Buffalo,New York,USA. {pkr, govind} @cedar.buffalo.edu http://www.cubs.buffalo.edu

More information

J. Weston, A. Gammerman, M. Stitson, V. Vapnik, V. Vovk, C. Watkins. Technical Report. February 5, 1998

J. Weston, A. Gammerman, M. Stitson, V. Vapnik, V. Vovk, C. Watkins. Technical Report. February 5, 1998 Density Estimation using Support Vector Machines J. Weston, A. Gammerman, M. Stitson, V. Vapnik, V. Vovk, C. Watkins. Technical Report CSD-TR-97-3 February 5, 998!()+, -./ 3456 Department of Computer Science

More information

Biostatistics 615/815 Lecture 16: Importance sampling Single dimensional optimization

Biostatistics 615/815 Lecture 16: Importance sampling Single dimensional optimization Biostatistics 615/815 Lecture 16: Single dimensional optimization Hyun Min Kang November 1st, 2012 Hyun Min Kang Biostatistics 615/815 - Lecture 16 November 1st, 2012 1 / 59 The crude Monte-Carlo Methods

More information

Week 10 Part A MIS 5214

Week 10 Part A MIS 5214 Week 10 Part A MIS 5214 Agenda Project Authentication Biometrics Access Control Models (DAC Part A) Access Control Techniques Centralized Remote Access Control Technologies Project assignment You and your

More information

STAT 113: Lab 9. Colin Reimer Dawson. Last revised November 10, 2015

STAT 113: Lab 9. Colin Reimer Dawson. Last revised November 10, 2015 STAT 113: Lab 9 Colin Reimer Dawson Last revised November 10, 2015 We will do some of the following together. The exercises with a (*) should be done and turned in as part of HW9. Before we start, let

More information

Adaptive Management of Multimodal Biometrics Fusion Using Ant Colony Optimization

Adaptive Management of Multimodal Biometrics Fusion Using Ant Colony Optimization Adaptive Management of Multimodal Biometrics Fusion Using Ant Colony Optimization Amioy Kumar, Ajay Kumar Corresponding author email: csajaykr@comp.polyu.edu.hk; Ph: 852-27667254 Abstract: This paper presents

More information

MODEL SELECTION AND REGULARIZATION PARAMETER CHOICE

MODEL SELECTION AND REGULARIZATION PARAMETER CHOICE MODEL SELECTION AND REGULARIZATION PARAMETER CHOICE REGULARIZATION METHODS FOR HIGH DIMENSIONAL LEARNING Francesca Odone and Lorenzo Rosasco odone@disi.unige.it - lrosasco@mit.edu June 6, 2011 ABOUT THIS

More information

Chapter 2 Describing, Exploring, and Comparing Data

Chapter 2 Describing, Exploring, and Comparing Data Slide 1 Chapter 2 Describing, Exploring, and Comparing Data Slide 2 2-1 Overview 2-2 Frequency Distributions 2-3 Visualizing Data 2-4 Measures of Center 2-5 Measures of Variation 2-6 Measures of Relative

More information

Edge and corner detection

Edge and corner detection Edge and corner detection Prof. Stricker Doz. G. Bleser Computer Vision: Object and People Tracking Goals Where is the information in an image? How is an object characterized? How can I find measurements

More information

Tutorial 1. Jun Xu, Teaching Asistant January 26, COMP4134 Biometrics Authentication

Tutorial 1. Jun Xu, Teaching Asistant January 26, COMP4134 Biometrics Authentication Tutorial 1 Jun Xu, Teaching Asistant csjunxu@comp.polyu.edu.hk COMP4134 Biometrics Authentication January 26, 2017 Table of Contents Problems Problem 1: Answer the following questions Problem 2: Biometric

More information

Chapter 3: User Authentication

Chapter 3: User Authentication Chapter 3: User Authentication Comp Sci 3600 Security Outline 1 2 3 4 Outline 1 2 3 4 User Authentication NIST SP 800-63-3 (Digital Authentication Guideline, October 2016) defines user as: The process

More information

Predictive Analysis: Evaluation and Experimentation. Heejun Kim

Predictive Analysis: Evaluation and Experimentation. Heejun Kim Predictive Analysis: Evaluation and Experimentation Heejun Kim June 19, 2018 Evaluation and Experimentation Evaluation Metrics Cross-Validation Significance Tests Evaluation Predictive analysis: training

More information

TRACK MAINTENANCE STRATEGIES OPTIMISATION PROBLEM

TRACK MAINTENANCE STRATEGIES OPTIMISATION PROBLEM TRACK MAINTENANCE STRATEGIES OPTIMISATION PROBLEM Gregory A. Krug Dr. S Krug Consulting Service P.O.B. 44051 Tel-Aviv 61440, Israel Viig@Inter.Net.Il Janusz Madejski Silesian University Of Technology In

More information

Multimodal Fusion Vulnerability to Non-Zero Effort (Spoof) Imposters

Multimodal Fusion Vulnerability to Non-Zero Effort (Spoof) Imposters Multimodal Fusion Vulnerability to Non-Zero Effort (Spoof) mposters P. A. Johnson, B. Tan, S. Schuckers 3 ECE Department, Clarkson University Potsdam, NY 3699, USA johnsopa@clarkson.edu tanb@clarkson.edu

More information

SD 372 Pattern Recognition

SD 372 Pattern Recognition SD 372 Pattern Recognition Lab 2: Model Estimation and Discriminant Functions 1 Purpose This lab examines the areas of statistical model estimation and classifier aggregation. Model estimation will be

More information

Using Support Vector Machines to Eliminate False Minutiae Matches during Fingerprint Verification

Using Support Vector Machines to Eliminate False Minutiae Matches during Fingerprint Verification Using Support Vector Machines to Eliminate False Minutiae Matches during Fingerprint Verification Abstract Praveer Mansukhani, Sergey Tulyakov, Venu Govindaraju Center for Unified Biometrics and Sensors

More information

Ground truth and evaluation for latent fingerprint matching

Ground truth and evaluation for latent fingerprint matching Ground truth and evaluation for latent fingerprint matching Anna Mikaelyan and Josef Bigun Halmstad University SE-0118 Halmstad {anna.mikaelyan,josef.bigun}@hh.se Abstract In forensic fingerprint studies

More information

Test Oracles and Randomness

Test Oracles and Randomness Test Oracles and Randomness UNIVERSITÄ T ULM DOCENDO CURANDO SCIENDO Ralph Guderlei and Johannes Mayer rjg@mathematik.uni-ulm.de, jmayer@mathematik.uni-ulm.de University of Ulm, Department of Stochastics

More information

Optimal Denoising of Natural Images and their Multiscale Geometry and Density

Optimal Denoising of Natural Images and their Multiscale Geometry and Density Optimal Denoising of Natural Images and their Multiscale Geometry and Density Department of Computer Science and Applied Mathematics Weizmann Institute of Science, Israel. Joint work with Anat Levin (WIS),

More information

Evaluating Classifiers

Evaluating Classifiers Evaluating Classifiers Reading for this topic: T. Fawcett, An introduction to ROC analysis, Sections 1-4, 7 (linked from class website) Evaluating Classifiers What we want: Classifier that best predicts

More information

Page 1. Graphical and Numerical Statistics

Page 1. Graphical and Numerical Statistics TOPIC: Description Statistics In this tutorial, we show how to use MINITAB to produce descriptive statistics, both graphical and numerical, for an existing MINITAB dataset. The example data come from Exercise

More information

AN ENHANCED METHOD FOR EXTREME LOADS ANALYSIS

AN ENHANCED METHOD FOR EXTREME LOADS ANALYSIS Brodogradnja/Shipbuilding/Open access Volume 68 Number 2, 2017 Francesco MAURO Radoslav NABERGOJ http://dx.doi.org/10.21278/brod68206 ISSN 0007-215X eissn 1845-5859 AN ENHANCED METHOD FOR EXTREME LOADS

More information

Statistics 202: Data Mining. c Jonathan Taylor. Outliers Based in part on slides from textbook, slides of Susan Holmes.

Statistics 202: Data Mining. c Jonathan Taylor. Outliers Based in part on slides from textbook, slides of Susan Holmes. Outliers Based in part on slides from textbook, slides of Susan Holmes December 2, 2012 1 / 1 Concepts What is an outlier? The set of data points that are considerably different than the remainder of the

More information

Stat 528 (Autumn 2008) Density Curves and the Normal Distribution. Measures of center and spread. Features of the normal distribution

Stat 528 (Autumn 2008) Density Curves and the Normal Distribution. Measures of center and spread. Features of the normal distribution Stat 528 (Autumn 2008) Density Curves and the Normal Distribution Reading: Section 1.3 Density curves An example: GRE scores Measures of center and spread The normal distribution Features of the normal

More information

Edge detection. Goal: Identify sudden. an image. Ideal: artist s line drawing. object-level knowledge)

Edge detection. Goal: Identify sudden. an image. Ideal: artist s line drawing. object-level knowledge) Edge detection Goal: Identify sudden changes (discontinuities) in an image Intuitively, most semantic and shape information from the image can be encoded in the edges More compact than pixels Ideal: artist

More information

Moving Beyond Linearity

Moving Beyond Linearity Moving Beyond Linearity Basic non-linear models one input feature: polynomial regression step functions splines smoothing splines local regression. more features: generalized additive models. Polynomial

More information

MODEL SELECTION AND REGULARIZATION PARAMETER CHOICE

MODEL SELECTION AND REGULARIZATION PARAMETER CHOICE MODEL SELECTION AND REGULARIZATION PARAMETER CHOICE REGULARIZATION METHODS FOR HIGH DIMENSIONAL LEARNING Francesca Odone and Lorenzo Rosasco odone@disi.unige.it - lrosasco@mit.edu June 3, 2013 ABOUT THIS

More information