ECEN 689 Special Topics in Data Science for Communications Networks

Size: px
Start display at page:

Download "ECEN 689 Special Topics in Data Science for Communications Networks"

Transcription

1 ECEN 689 Special Topics in Data Science for Communications Networks Nick Duffield Department of Electrical & Computer Engineering Texas A&M University Lecture 13 Measuring and Inferring Traffic Matrices

2 Measuring and Inferring Network Usage Flow records à Origin Destination Traffic Matrices What if we don t have collect flow records network wide Core backbone network: yes Access networks: sometimes not

3 Sources of ISP Operational Data Router Centers Peering Access Backbone Business Management Datacenters

4 Sources of ISP Operational Data Traffic flow through core: NetFlow capture in core routers

5 Sources of ISP Operational Data Traffic flow in access network: does not touch core: no NetFlow

6 Sources of ISP Operational Data Router Centers Peering Access Backbone Link Traffic Rates Timeseries of traffic per router interface, 5 minute granularity Business 0:00 0:05 0:10 0:15 0:20 0:25 0:30 0:35 Management Datacenters

7 Router Management Information Base Management Information Base (MIB) Database used for managing routers and other network devices Hierarchical database on managed objects maintained in router E.g. operational statistics, interface parameters and settings Interact with MIB via SNMP Simple Network Management Protocol Operates on a pull model Poll device using SNMP to retrieve statistics Contrast push model of NetFlow: export records to a collector Actually, can set traps Event contingencies that generate actions if they occur When interface goes down, send warning to Doesn t scale to anything resembling NetFlow

8 Router Interface Counters Managed objects in table maintained in MIB, per interface Byte and packet counters 64 bit counters; 32 bit can wrap too quickly Poll interfaces periodically Time series of counter value Average rate = CounterDifference / PollingInterval Time series of average rates 5 minute polling interval is common practice Conservative, historical concerns about load on router Data issues Round robin polling of routers during period: intervals don t align Counter rollover detection and correction Detect and flag missing data values Router did not respond to SNMP poll, network congestion / outage

9 Packet level traffic measurements The other end of the spectrum! Special purpose platforms collecting packet header traces Enables detailed statistical analysis of packet arrival Help develop traffic models for use in analysis

10 Poisson Process increasing window width t Poisson(λ) Poisson process with intensity λ Interarrival times IID exponential Mean interarrival time 1/λ Arrivals in time window t Poisson RV: mean = variance = λt Coefficient of variation Mean/SD = (λt) -1/2 Process becomes smoother as window width t increases Good model for telephone call start process Time-of-day rate dependence

11 Actual Packet Arrival Process increasing window width t Arrivals do not smooth out with increasing time window Variability at all timescales Self-similarity Process looks identical at all timescale Example of self-similar process Fractional Brownian motion (fbm) Window scaling by factor a B H (a t) = a H/2 B H (t) some H > 0 H = ½ is standard Brownian motion H > ½ positive correlations Long range correlations give rise to burstiness at all time scales

12 Traffic matrices from link measurements? Y 1 Y n+1 C Y 2n Y n Know link aggregate rates into C: (Y 1,,Y n ) and out of C: (Y n+1,,y 2n ) Want traffic matrices X ij = traffic from i to j

13 Linear Relations Y i : traffic in from i, Y n+j : traffic out to j; X ij : traffic from i to j Speak of ij as origin-destination (OD) pair What do X and Y model? First model: Y, as average rates over the same single measurement interval Y i = Σ j X ij, i = 1,,n traffic from origin i = sum over destinations j of traffic from i to j Y n+j = Σ i X ij j=1,,n traffic to destination j = sum over origins i of traffic from i to j Linear relation Y = A.X A is routing matrix A lp = 1 if traffic on OD pair p is carried over link l, 0 otherwise

14 Can we solve Y = A X? X ij : n 2 OD rates from Y i: 2n link rates? E.g. n = 3: 9 OD rates, 6 link rates X=( X 11, X 12, X 13, X 21, X 22, X 23, X 31, X 32, X 33 ) Y=( Y 1, Y 2, Y 3, Y 4, Y 5, Y 6 ): 1,2,3 in, 4,5,6 out A = Underconstrained linear system A is not of full rank only 5 linearly independent rows/columns

15 Statistical Modeling Make a statistical model of OD traffic Specified by some parameter set ϕ Link traffic is composite of OD that traverses it Analysis to find composite parameters θ = F(ϕ) of link traffic Suppose we could determine the link parameters θ Measure, estimate θ If F is invertible then we could estimate ϕ = F -1 (θ ) Is F invertible? are the OD parameters ϕ uniquely identifiable from link measurements?

16 Simple Poisson Model Model OD traffic as independent Poisson processes Traffic for OD pair ij is Poisson(X ij ) X ij is intensity, i.e. mean arrival rate per unit time Inter-arrival times IID, exponential distribution, mean 1/X ij What is the composite model for link traffic Additivity Poisson(X) + Poisson(X ) = d Poisson(X+X ) Link processes Link i process is Poisson with intensity Y i = (A.X) j Invertible? (Find X uniquely in terms of Y?) Same problem as before: under-constrained linear system

17 Exploit whole random process? Consider {x ij (t): t=1,2,..,t} n values of OD rates during T successive time periods Independent random processes Corresponding link processes: y i (t) = (A. x(t)) i How to get n 2 known quantities? Covariances amongst the n 2 pairs of inbound and outbound traffic Inbound y i (t) = Σ m x im (t), i = 1,,n Outbound y n+j (t) = Σ k x kj (t), j = 1,,n Cov( y i, y n+j ) = Σ mk Cov(x im, x kj ) = Var(x ij ) Assume independent OD processes: only nonzero terms are m=j, k=i Recover variance of OD processes Later: how to generalize this approach

18 General Setting Origin-Destination (OD) Pairs: j {1,,c); Poisson processes of intensity λ j; Observed values X j (t) in time period t=1,,t Directed Links i (1,,r) Routing matrix A ij = 1 if OD pair j routed over link i Link processes Y= AX Y i (t) = Σ j A ij X j (t) Poissonian with rates = Σ j A ij λ j; When can the { λ j } be identified from the processes Y Identified from the values {Y i (t): i=1,,r; t=1,,t}

19 Identifiability and Estimation Identifiability Distinct sets of OD rates { λ j } give distinct distributions of Y = { Y i (t) } Estimation If { λ j } identifiable A mapping from the outcomes Y = { Y i (t) } to estimates { λ j } of { λ j } Good statistical properties desirable Some measure of accuracy, e.g. unbiased, variance, Computable in useful cases Complexity feasible for use in real networks

20 When is general Poissonian model identifiable? Theorem: if routing matrix A Has distinct columns Each column has at least one non-zero entry Then the { λ j } are identifiable Necessity If columns j and k identical, Y depends only on sum (X j + X k ) Could only identify the sum λ j + λ k, not the constituents If column j zero, Y does not depend on X j at all Sufficiency See Vardi 1996

21 Moment-based Estimator Match model mean and covariance to sample versions Normal approxination Mean: E[Y] = A λ Sample mean: Y = t Y(t) / T Covariance: Cov(Y j, Y k ) = = il A ji. cov(x i, X l ) A kl = (A. diag(λ). A T ) ij Sample covariance: S jk = t (Y j (t)y k (t) Y j Y k ) / T

22 Linear Estimation Equate sample and model versions Y = Aλ and S = Bλ where B jk,i = A ji A ik Back to solving linear equation for λ Difficulties System generally inconsistent if S ii (Y i ) 2 System generally over-constrained Approach A = r x c, B = r(r-1) x c Iterative approach LinInPos (Linear Inversion with Positive Constraints) Some success in example cases

23 Maximum Likelihood Estimator Likelihood function: L(λ) = Pr λ (Y) = X: Y=AX Pr λ (X) Maximum likelihood estimate Find MLE λ that maximizes L(λ) Often difficult to compute Difficult to determine set {X: Y=AX} MLE may lie on boundary of {X : Y= AX } Iterative methods Expectation-Maximization Algorithm Given value λ, find λ* that maximizes E X Y,λ* [ log L(λ) ] Iterative mapping λ à λ* converge (in some cases) to MLE But not always

24 Time-dependent model (Cao 2000) OD process: X = Normal(λ t,σ t ): λ =(λ 1,,λ n ), σ t = covariance matrix Link process: Y t = A.X t = Normal( Aλ t, Aσ t A T ) Constraint: σ t = k * diag(λ tc ) Some constant k > 0 and exponent c > 1 Exponent c > 1 captures higher variance that Poisson Simple characterization inspired by network traffic measurement MLE via EM High computation cost O(r 5 )

25 Can structural models help? Gravity model Assume product form for OD traffic Estimate X ij by X ij G = k Y in i * Yout j k is constant chosen for correct normalization of total traffic Gravity model by analogy to Newton s Law of Gravitation Attractive force between two objects proportional to product of masses How good is the gravity model? Computation Very quick to compute Accuracy X ij G does not solve the linear system Y = A.X in general Need to investigate experimentally

26 Gravity Model Performance Not very accurate, although some improvement possible by aggregation in real networks

27 Generalized Gravity Model Internet routing is asymmetric Hot potato routing: use the closest exit point Generalized gravity model For outbound traffic, assumes proportionality on per-peer basis as opposed to per-router peer links access links

28 Generalized Gravity Model

29 Can we improve on the Gravity Model? For gravity model : quick to compute Against gravity model: does not satisfy constraint Y = AX Can we get a best of both worlds approach? Compromise between Gravity Model and Constraint?

30 Tomogravity Tomogravity = tomography + gravity modeling Use least-squares method to get the solution, which Satisfies the constraints Y = A.X Is closest to the gravity model solution Can use weighted least-squares to make more robust least square solution gravity model solution constraint subspace Y = A.X

31 Tomogravity: Accuracy Accurate within 10-20% (esp. for large elements)

32 Regularization Want a solution that satisfies constraints: Y = AX Many more unknowns than measurement: N end points: O(N 2 ) paths, O(N) links Underconstrained linear system Many solutions satisfy the equations Must somehow choose the best solution Regularization Include penalty function J(X) J(X) expresses preference amongst different possible solutions Estimate X : X = argmin X { Y- AX 2 + µ 2 J(X) }

33 Regularization in Tomogravity Consider J(X) = X X G 2 Norm distance from Gravity Model Solution X = argmin X { µ -2 Y- AX 2 + X X G 2 } For small µ, X approaches least squares solution argmin X: Y- AX { X X G 2 } More robust results found with weighted least squares solution J(X) = (X X G ) / X G 2 component-wise division by X G

34 Information and Entropy X with discrete distribution {p(x i ), i=1,..,n}} Entropy: H(X) = - Σ i p(x i ) log p(x i ) H(X) = 0 if one p(x i ) = 1 (all others 0) H(X) = n -1 log n id all p(x i )= 1/n : max value H(X) a measure of uncertainty of X - p log p

35 Mutual Information Entropy: H(X) = - Σ i p(x i ) log p(x i ) Conditional entropy H(Z X) X-averaged uncertainty in conditional distribution of Z X H(Z X) = - Σ j p (x j ) Σ j p (z i x j ) log Σ j p (z i x j ) Note if Z and X are independent p (z i x j ) = p (z i ) : H(Z X) = H(Z) Mutual information I(Z,X) = H(Z) H(Z X) I(Z,X) 0 with equality iff Z, X are independent I(Z,X) = Σ ij p (x i, z i ) log p (x i, z i ) / p (x i ) p (z i ) Independent case: p (x i, z i ) = p (x i ) p (z i ): I(Z,X) = 0

36 Traffic Matrices and Information What does the distribution of source traffic tell us about the distribution at the destinations First: normalize the OD matrix p(s,d) = fraction of all traffic from source s to destination d Joint distribution of source and destination traffic p S (d) = Σ s p(s,d): fraction of traffic to destination d p D (s) = Σ d p(s,d): fraction of traffic from source s Suppose p(s,d) = p D (s) * p S (d) Source and destination distributions independent: GRAVITY MODEL Mutual information I(S,D) = Σ s,d p (s, d ) log p (s, d ) / p D (s) p S (d) How much source distribution tells us about destination distribution Gravity model: no mutual information I(S,D) = 0

37 Regularization and MMI MMI: minimal mutual information Regularization with penalty function J = I(S,D) X = argmin X { Y- AX 2 +µ 2 I(S,D) } Intuition: gravity model represents lack of information No model of source / destination dependence Find compromise between that and constraints Y = AX Relation to original tomogravity For close distributions P and Q I(P,Q) Σ i (p i q i ) / q i 2 Weighted mean square distance from before

38 References Network Tomography: Estimating Source-Destination Traffic Intensities From Link Data, Vardi, Time varying network tomography: router link data, Cao et. al, Estimating Point-to-Point and Point-to-Multipoint Traffic Matrices: An Information-Theoretic Approach, Zhang, Roughan, Lund & Donoho, Chapter 9 of Kolaczyk: Statistical Analysis of Network Data: Methods and Models.

An Overview of Traffic Matrix Estimation Methods

An Overview of Traffic Matrix Estimation Methods An Overview of Traffic Matrix Estimation Methods Nina Taft Berkeley www.intel.com/research Problem Statement 1 st generation solutions 2 nd generation solutions 3 rd generation solutions Summary Outline

More information

Ph.D. Qualifying Exam Friday Saturday, January 6 7, 2017

Ph.D. Qualifying Exam Friday Saturday, January 6 7, 2017 Ph.D. Qualifying Exam Friday Saturday, January 6 7, 2017 Put your solution to each problem on a separate sheet of paper. Problem 1. (5106) Let X 1, X 2,, X n be a sequence of i.i.d. observations from a

More information

Network Performance Tomography

Network Performance Tomography Network Performance Tomography Hung X. Nguyen TeleTraffic Research Center University of Adelaide Network Performance Tomography Inferring link performance using end-to-end probes 2 Network Performance

More information

ECEN 689 Special Topics in Data Science for Communications Networks

ECEN 689 Special Topics in Data Science for Communications Networks ECEN 689 Special Topics in Data Science for Communications Networks Nick Duffield Department of Electrical & Computer Engineering Texas A&M University Lecture 3 Estimation and Bounds Estimation from Packet

More information

Learning the Linear Dynamical System with ASOS ( Approximated Second-Order Statistics )

Learning the Linear Dynamical System with ASOS ( Approximated Second-Order Statistics ) Learning the Linear Dynamical System with ASOS ( Approximated Second-Order Statistics ) James Martens University of Toronto June 24, 2010 Computer Science UNIVERSITY OF TORONTO James Martens (U of T) Learning

More information

PROBABILITY AND INFORMATION THEORY. Dr. Gjergji Kasneci Introduction to Information Retrieval WS

PROBABILITY AND INFORMATION THEORY. Dr. Gjergji Kasneci Introduction to Information Retrieval WS PROBABILITY AND INFORMATION THEORY Dr. Gjergji Kasneci Introduction to Information Retrieval WS 2012-13 1 Outline Intro Basics of probability and information theory Probability space Rules of probability

More information

Notes on Random Vectors and Multivariate Normal

Notes on Random Vectors and Multivariate Normal MATH 590 Spring 06 Notes on Random Vectors and Multivariate Normal Properties of Random Vectors If X,, X n are random variables, then X = X,, X n ) is a random vector, with the cumulative distribution

More information

Maximum Likelihood Estimation of the Flow Size Distribution Tail Index from Sampled Packet Data

Maximum Likelihood Estimation of the Flow Size Distribution Tail Index from Sampled Packet Data Maximum Likelihood Estimation of the Flow Size Distribution Tail Index from Sampled Packet Data Patrick Loiseau 1, Paulo Gonçalves 1, Stéphane Girard 2, Florence Forbes 2, Pascale Vicat-Blanc Primet 1

More information

Likelihood, MLE & EM for Gaussian Mixture Clustering. Nick Duffield Texas A&M University

Likelihood, MLE & EM for Gaussian Mixture Clustering. Nick Duffield Texas A&M University Likelihood, MLE & EM for Gaussian Mixture Clustering Nick Duffield Texas A&M University Probability vs. Likelihood Probability: predict unknown outcomes based on known parameters: P(x q) Likelihood: estimate

More information

The Multivariate Gaussian Distribution [DRAFT]

The Multivariate Gaussian Distribution [DRAFT] The Multivariate Gaussian Distribution DRAFT David S. Rosenberg Abstract This is a collection of a few key and standard results about multivariate Gaussian distributions. I have not included many proofs,

More information

Name: Matriculation Number: Tutorial Group: A B C D E

Name: Matriculation Number: Tutorial Group: A B C D E Name: Matriculation Number: Tutorial Group: A B C D E Question: 1 (5 Points) 2 (6 Points) 3 (5 Points) 4 (5 Points) Total (21 points) Score: General instructions: The written test contains 4 questions

More information

Lecture 4: Sampling, Tail Inequalities

Lecture 4: Sampling, Tail Inequalities Lecture 4: Sampling, Tail Inequalities Variance and Covariance Moment and Deviation Concentration and Tail Inequalities Sampling and Estimation c Hung Q. Ngo (SUNY at Buffalo) CSE 694 A Fun Course 1 /

More information

Review (Probability & Linear Algebra)

Review (Probability & Linear Algebra) Review (Probability & Linear Algebra) CE-725 : Statistical Pattern Recognition Sharif University of Technology Spring 2013 M. Soleymani Outline Axioms of probability theory Conditional probability, Joint

More information

cs/ee/ids 143 Communication Networks

cs/ee/ids 143 Communication Networks cs/ee/ids 143 Communication Networks Chapter 5 Routing Text: Walrand & Parakh, 2010 Steven Low CMS, EE, Caltech Warning These notes are not self-contained, probably not understandable, unless you also

More information

An Independent-Connection Model for Traffic Matrices

An Independent-Connection Model for Traffic Matrices An Independent-Connection Model for Traffic Matrices Vijay Erramilli, Mark Crovella, and Nina Taft Λ Sept. 7 26 BUCS-TR-26-22 Abstract A common assumption made in traffic matrix (TM) modeling and estimation

More information

Machine Learning for Large-Scale Data Analysis and Decision Making A. Week #1

Machine Learning for Large-Scale Data Analysis and Decision Making A. Week #1 Machine Learning for Large-Scale Data Analysis and Decision Making 80-629-17A Week #1 Today Introduction to machine learning The course (syllabus) Math review (probability + linear algebra) The future

More information

CS 246 Review of Linear Algebra 01/17/19

CS 246 Review of Linear Algebra 01/17/19 1 Linear algebra In this section we will discuss vectors and matrices. We denote the (i, j)th entry of a matrix A as A ij, and the ith entry of a vector as v i. 1.1 Vectors and vector operations A vector

More information

EE263 Review Session 1

EE263 Review Session 1 EE263 Review Session 1 October 5, 2018 0.1 Importing Variables from a MALAB.m file If you are importing variables given in file vars.m, use the following code at the beginning of your script. close a l

More information

Charging from Sampled Network Usage

Charging from Sampled Network Usage Charging from Sampled Network Usage Nick Duffield Carsten Lund Mikkel Thorup AT&T Labs-Research, Florham Park, NJ 1 Do Charging and Sampling Mix? Usage sensitive charging charge based on sampled network

More information

Lecture: Examples of LP, SOCP and SDP

Lecture: Examples of LP, SOCP and SDP 1/34 Lecture: Examples of LP, SOCP and SDP Zaiwen Wen Beijing International Center For Mathematical Research Peking University http://bicmr.pku.edu.cn/~wenzw/bigdata2018.html wenzw@pku.edu.cn Acknowledgement:

More information

STAT 100C: Linear models

STAT 100C: Linear models STAT 100C: Linear models Arash A. Amini April 27, 2018 1 / 1 Table of Contents 2 / 1 Linear Algebra Review Read 3.1 and 3.2 from text. 1. Fundamental subspace (rank-nullity, etc.) Im(X ) = ker(x T ) R

More information

Efficient MCMC Samplers for Network Tomography

Efficient MCMC Samplers for Network Tomography Efficient MCMC Samplers for Network Tomography Martin Hazelton 1 Institute of Fundamental Sciences Massey University 7 December 2015 1 Email: m.hazelton@massey.ac.nz AUT Mathematical Sciences Symposium

More information

Mathematical statistics

Mathematical statistics October 1 st, 2018 Lecture 11: Sufficient statistic Where are we? Week 1 Week 2 Week 4 Week 7 Week 10 Week 14 Probability reviews Chapter 6: Statistics and Sampling Distributions Chapter 7: Point Estimation

More information

Intro Refresher Reversibility Open networks Closed networks Multiclass networks Other networks. Queuing Networks. Florence Perronnin

Intro Refresher Reversibility Open networks Closed networks Multiclass networks Other networks. Queuing Networks. Florence Perronnin Queuing Networks Florence Perronnin Polytech Grenoble - UGA March 23, 27 F. Perronnin (UGA) Queuing Networks March 23, 27 / 46 Outline Introduction to Queuing Networks 2 Refresher: M/M/ queue 3 Reversibility

More information

STATS 306B: Unsupervised Learning Spring Lecture 12 May 7

STATS 306B: Unsupervised Learning Spring Lecture 12 May 7 STATS 306B: Unsupervised Learning Spring 2014 Lecture 12 May 7 Lecturer: Lester Mackey Scribe: Lan Huong, Snigdha Panigrahi 12.1 Beyond Linear State Space Modeling Last lecture we completed our discussion

More information

Sampling and incomplete network data

Sampling and incomplete network data 1/58 Sampling and incomplete network data 567 Statistical analysis of social networks Peter Hoff Statistics, University of Washington 2/58 Network sampling methods It is sometimes difficult to obtain a

More information

SPLITTING AND MERGING OF PACKET TRAFFIC: MEASUREMENT AND MODELLING

SPLITTING AND MERGING OF PACKET TRAFFIC: MEASUREMENT AND MODELLING SPLITTING AND MERGING OF PACKET TRAFFIC: MEASUREMENT AND MODELLING Nicolas Hohn 1 Darryl Veitch 1 Tao Ye 2 1 CUBIN, Department of Electrical & Electronic Engineering University of Melbourne, Vic 3010 Australia

More information

Linear Regression and Its Applications

Linear Regression and Its Applications Linear Regression and Its Applications Predrag Radivojac October 13, 2014 Given a data set D = {(x i, y i )} n the objective is to learn the relationship between features and the target. We usually start

More information

Notes on the Multivariate Normal and Related Topics

Notes on the Multivariate Normal and Related Topics Version: July 10, 2013 Notes on the Multivariate Normal and Related Topics Let me refresh your memory about the distinctions between population and sample; parameters and statistics; population distributions

More information

Designing Information Devices and Systems I Fall 2018 Lecture Notes Note 6

Designing Information Devices and Systems I Fall 2018 Lecture Notes Note 6 EECS 16A Designing Information Devices and Systems I Fall 2018 Lecture Notes Note 6 6.1 Introduction: Matrix Inversion In the last note, we considered a system of pumps and reservoirs where the water in

More information

ELE539A: Optimization of Communication Systems Lecture 6: Quadratic Programming, Geometric Programming, and Applications

ELE539A: Optimization of Communication Systems Lecture 6: Quadratic Programming, Geometric Programming, and Applications ELE539A: Optimization of Communication Systems Lecture 6: Quadratic Programming, Geometric Programming, and Applications Professor M. Chiang Electrical Engineering Department, Princeton University February

More information

Network Anomography. Abstract

Network Anomography. Abstract etwork Anomography Yin Zhang, Zihui Ge, Albert Greenberg, and Matthew Roughan Department of Computer Sciences, University of Texas at Austin, Austin, TX 78712, USA AT&T Labs Research, Florham Park, J 7932,

More information

ECE353: Probability and Random Processes. Lecture 18 - Stochastic Processes

ECE353: Probability and Random Processes. Lecture 18 - Stochastic Processes ECE353: Probability and Random Processes Lecture 18 - Stochastic Processes Xiao Fu School of Electrical Engineering and Computer Science Oregon State University E-mail: xiao.fu@oregonstate.edu From RV

More information

Stochastic process. X, a series of random variables indexed by t

Stochastic process. X, a series of random variables indexed by t Stochastic process X, a series of random variables indexed by t X={X(t), t 0} is a continuous time stochastic process X={X(t), t=0,1, } is a discrete time stochastic process X(t) is the state at time t,

More information

A Introduction to Matrix Algebra and the Multivariate Normal Distribution

A Introduction to Matrix Algebra and the Multivariate Normal Distribution A Introduction to Matrix Algebra and the Multivariate Normal Distribution PRE 905: Multivariate Analysis Spring 2014 Lecture 6 PRE 905: Lecture 7 Matrix Algebra and the MVN Distribution Today s Class An

More information

EXAMINATIONS OF THE HONG KONG STATISTICAL SOCIETY

EXAMINATIONS OF THE HONG KONG STATISTICAL SOCIETY EXAMINATIONS OF THE HONG KONG STATISTICAL SOCIETY HIGHER CERTIFICATE IN STATISTICS, 2013 MODULE 5 : Further probability and inference Time allowed: One and a half hours Candidates should answer THREE questions.

More information

COUNTING INTEGER POINTS IN POLYHEDRA. Alexander Barvinok

COUNTING INTEGER POINTS IN POLYHEDRA. Alexander Barvinok COUNTING INTEGER POINTS IN POLYHEDRA Alexander Barvinok Papers are available at http://www.math.lsa.umich.edu/ barvinok/papers.html Let P R d be a polytope. We want to compute (exactly or approximately)

More information

MA 575 Linear Models: Cedric E. Ginestet, Boston University Revision: Probability and Linear Algebra Week 1, Lecture 2

MA 575 Linear Models: Cedric E. Ginestet, Boston University Revision: Probability and Linear Algebra Week 1, Lecture 2 MA 575 Linear Models: Cedric E Ginestet, Boston University Revision: Probability and Linear Algebra Week 1, Lecture 2 1 Revision: Probability Theory 11 Random Variables A real-valued random variable is

More information

1 Outline. 1. Motivation. 2. SUR model. 3. Simultaneous equations. 4. Estimation

1 Outline. 1. Motivation. 2. SUR model. 3. Simultaneous equations. 4. Estimation 1 Outline. 1. Motivation 2. SUR model 3. Simultaneous equations 4. Estimation 2 Motivation. In this chapter, we will study simultaneous systems of econometric equations. Systems of simultaneous equations

More information

(y 1, y 2 ) = 12 y3 1e y 1 y 2 /2, y 1 > 0, y 2 > 0 0, otherwise.

(y 1, y 2 ) = 12 y3 1e y 1 y 2 /2, y 1 > 0, y 2 > 0 0, otherwise. 54 We are given the marginal pdfs of Y and Y You should note that Y gamma(4, Y exponential( E(Y = 4, V (Y = 4, E(Y =, and V (Y = 4 (a With U = Y Y, we have E(U = E(Y Y = E(Y E(Y = 4 = (b Because Y and

More information

TAMS39 Lecture 2 Multivariate normal distribution

TAMS39 Lecture 2 Multivariate normal distribution TAMS39 Lecture 2 Multivariate normal distribution Martin Singull Department of Mathematics Mathematical Statistics Linköping University, Sweden Content Lecture Random vectors Multivariate normal distribution

More information

Part IB Statistics. Theorems with proof. Based on lectures by D. Spiegelhalter Notes taken by Dexter Chua. Lent 2015

Part IB Statistics. Theorems with proof. Based on lectures by D. Spiegelhalter Notes taken by Dexter Chua. Lent 2015 Part IB Statistics Theorems with proof Based on lectures by D. Spiegelhalter Notes taken by Dexter Chua Lent 2015 These notes are not endorsed by the lecturers, and I have modified them (often significantly)

More information

Modeling Residual-Geometric Flow Sampling

Modeling Residual-Geometric Flow Sampling Modeling Residual-Geometric Flow Sampling Xiaoming Wang Joint work with Xiaoyong Li and Dmitri Loguinov Amazon.com Inc., Seattle, WA April 13 th, 2011 1 Agenda Introduction Underlying model of residual

More information

1. Let A be a 2 2 nonzero real matrix. Which of the following is true?

1. Let A be a 2 2 nonzero real matrix. Which of the following is true? 1. Let A be a 2 2 nonzero real matrix. Which of the following is true? (A) A has a nonzero eigenvalue. (B) A 2 has at least one positive entry. (C) trace (A 2 ) is positive. (D) All entries of A 2 cannot

More information

b. ( ) ( ) ( ) ( ) ( ) 5. Independence: Two events (A & B) are independent if one of the conditions listed below is satisfied; ( ) ( ) ( )

b. ( ) ( ) ( ) ( ) ( ) 5. Independence: Two events (A & B) are independent if one of the conditions listed below is satisfied; ( ) ( ) ( ) 1. Set a. b. 2. Definitions a. Random Experiment: An experiment that can result in different outcomes, even though it is performed under the same conditions and in the same manner. b. Sample Space: This

More information

Linear Algebra Methods for Data Mining

Linear Algebra Methods for Data Mining Linear Algebra Methods for Data Mining Saara Hyvönen, Saara.Hyvonen@cs.helsinki.fi Spring 2007 1. Basic Linear Algebra Linear Algebra Methods for Data Mining, Spring 2007, University of Helsinki Example

More information

11 a 12 a 21 a 11 a 22 a 12 a 21. (C.11) A = The determinant of a product of two matrices is given by AB = A B 1 1 = (C.13) and similarly.

11 a 12 a 21 a 11 a 22 a 12 a 21. (C.11) A = The determinant of a product of two matrices is given by AB = A B 1 1 = (C.13) and similarly. C PROPERTIES OF MATRICES 697 to whether the permutation i 1 i 2 i N is even or odd, respectively Note that I =1 Thus, for a 2 2 matrix, the determinant takes the form A = a 11 a 12 = a a 21 a 11 a 22 a

More information

Review (probability, linear algebra) CE-717 : Machine Learning Sharif University of Technology

Review (probability, linear algebra) CE-717 : Machine Learning Sharif University of Technology Review (probability, linear algebra) CE-717 : Machine Learning Sharif University of Technology M. Soleymani Fall 2012 Some slides have been adopted from Prof. H.R. Rabiee s and also Prof. R. Gutierrez-Osuna

More information

6.207/14.15: Networks Lectures 4, 5 & 6: Linear Dynamics, Markov Chains, Centralities

6.207/14.15: Networks Lectures 4, 5 & 6: Linear Dynamics, Markov Chains, Centralities 6.207/14.15: Networks Lectures 4, 5 & 6: Linear Dynamics, Markov Chains, Centralities 1 Outline Outline Dynamical systems. Linear and Non-linear. Convergence. Linear algebra and Lyapunov functions. Markov

More information

CS418 Operating Systems

CS418 Operating Systems CS418 Operating Systems Lecture 14 Queuing Analysis Textbook: Operating Systems by William Stallings 1 1. Why Queuing Analysis? If the system environment changes (like the number of users is doubled),

More information

Basic Concepts in Matrix Algebra

Basic Concepts in Matrix Algebra Basic Concepts in Matrix Algebra An column array of p elements is called a vector of dimension p and is written as x p 1 = x 1 x 2. x p. The transpose of the column vector x p 1 is row vector x = [x 1

More information

Machine learning - HT Maximum Likelihood

Machine learning - HT Maximum Likelihood Machine learning - HT 2016 3. Maximum Likelihood Varun Kanade University of Oxford January 27, 2016 Outline Probabilistic Framework Formulate linear regression in the language of probability Introduce

More information

Routing. Topics: 6.976/ESD.937 1

Routing. Topics: 6.976/ESD.937 1 Routing Topics: Definition Architecture for routing data plane algorithm Current routing algorithm control plane algorithm Optimal routing algorithm known algorithms and implementation issues new solution

More information

CS 542G: Robustifying Newton, Constraints, Nonlinear Least Squares

CS 542G: Robustifying Newton, Constraints, Nonlinear Least Squares CS 542G: Robustifying Newton, Constraints, Nonlinear Least Squares Robert Bridson October 29, 2008 1 Hessian Problems in Newton Last time we fixed one of plain Newton s problems by introducing line search

More information

Bounded Delay for Weighted Round Robin with Burst Crediting

Bounded Delay for Weighted Round Robin with Burst Crediting Bounded Delay for Weighted Round Robin with Burst Crediting Sponsor: Sprint Kert Mezger David W. Petr Technical Report TISL-0230-08 Telecommunications and Information Sciences Laboratory Department of

More information

UC Berkeley Department of Electrical Engineering and Computer Sciences. EECS 126: Probability and Random Processes

UC Berkeley Department of Electrical Engineering and Computer Sciences. EECS 126: Probability and Random Processes UC Berkeley Department of Electrical Engineering and Computer Sciences EECS 6: Probability and Random Processes Problem Set 3 Spring 9 Self-Graded Scores Due: February 8, 9 Submit your self-graded scores

More information

CS 591, Lecture 2 Data Analytics: Theory and Applications Boston University

CS 591, Lecture 2 Data Analytics: Theory and Applications Boston University CS 591, Lecture 2 Data Analytics: Theory and Applications Boston University Charalampos E. Tsourakakis January 25rd, 2017 Probability Theory The theory of probability is a system for making better guesses.

More information

Clustering and Gaussian Mixture Models

Clustering and Gaussian Mixture Models Clustering and Gaussian Mixture Models Piyush Rai IIT Kanpur Probabilistic Machine Learning (CS772A) Jan 25, 2016 Probabilistic Machine Learning (CS772A) Clustering and Gaussian Mixture Models 1 Recap

More information

Chapter 3, 4 Random Variables ENCS Probability and Stochastic Processes. Concordia University

Chapter 3, 4 Random Variables ENCS Probability and Stochastic Processes. Concordia University Chapter 3, 4 Random Variables ENCS6161 - Probability and Stochastic Processes Concordia University ENCS6161 p.1/47 The Notion of a Random Variable A random variable X is a function that assigns a real

More information

EXPECTED VALUE of a RV. corresponds to the average value one would get for the RV when repeating the experiment, =0.

EXPECTED VALUE of a RV. corresponds to the average value one would get for the RV when repeating the experiment, =0. EXPECTED VALUE of a RV corresponds to the average value one would get for the RV when repeating the experiment, independently, infinitely many times. Sample (RIS) of n values of X (e.g. More accurately,

More information

Optimal Combination of Sampled Network Measurements

Optimal Combination of Sampled Network Measurements Optimal Combination of Sampled Network Measurements Nick Duffield Carsten Lund Mikkel Thorup AT&T Labs Research, 8 Park Avenue, Florham Park, New Jersey, 7932, USA {duffield,lund,mthorup}research.att.com

More information

Generative and Discriminative Approaches to Graphical Models CMSC Topics in AI

Generative and Discriminative Approaches to Graphical Models CMSC Topics in AI Generative and Discriminative Approaches to Graphical Models CMSC 35900 Topics in AI Lecture 2 Yasemin Altun January 26, 2007 Review of Inference on Graphical Models Elimination algorithm finds single

More information

Machine Learning for Signal Processing Bayes Classification and Regression

Machine Learning for Signal Processing Bayes Classification and Regression Machine Learning for Signal Processing Bayes Classification and Regression Instructor: Bhiksha Raj 11755/18797 1 Recap: KNN A very effective and simple way of performing classification Simple model: For

More information

TWO PROBLEMS IN NETWORK PROBING

TWO PROBLEMS IN NETWORK PROBING TWO PROBLEMS IN NETWORK PROBING DARRYL VEITCH The University of Melbourne 1 Temporal Loss and Delay Tomography 2 Optimal Probing in Convex Networks Paris Networking 27 Juin 2007 TEMPORAL LOSS AND DELAY

More information

Lecture 1 Review: Linear models have the form (in matrix notation) Y = Xβ + ε,

Lecture 1 Review: Linear models have the form (in matrix notation) Y = Xβ + ε, 2. REVIEW OF LINEAR ALGEBRA 1 Lecture 1 Review: Linear models have the form (in matrix notation) Y = Xβ + ε, where Y n 1 response vector and X n p is the model matrix (or design matrix ) with one row for

More information

Clustering with k-means and Gaussian mixture distributions

Clustering with k-means and Gaussian mixture distributions Clustering with k-means and Gaussian mixture distributions Machine Learning and Object Recognition 2017-2018 Jakob Verbeek Clustering Finding a group structure in the data Data in one cluster similar to

More information

[y i α βx i ] 2 (2) Q = i=1

[y i α βx i ] 2 (2) Q = i=1 Least squares fits This section has no probability in it. There are no random variables. We are given n points (x i, y i ) and want to find the equation of the line that best fits them. We take the equation

More information

IE 581 Introduction to Stochastic Simulation

IE 581 Introduction to Stochastic Simulation 1. List criteria for choosing the majorizing density r (x) when creating an acceptance/rejection random-variate generator for a specified density function f (x). 2. Suppose the rate function of a nonhomogeneous

More information

CS281 Section 4: Factor Analysis and PCA

CS281 Section 4: Factor Analysis and PCA CS81 Section 4: Factor Analysis and PCA Scott Linderman At this point we have seen a variety of machine learning models, with a particular emphasis on models for supervised learning. In particular, we

More information

MIT Spring 2016

MIT Spring 2016 Generalized Linear Models MIT 18.655 Dr. Kempthorne Spring 2016 1 Outline Generalized Linear Models 1 Generalized Linear Models 2 Generalized Linear Model Data: (y i, x i ), i = 1,..., n where y i : response

More information

EE5139R: Problem Set 4 Assigned: 31/08/16, Due: 07/09/16

EE5139R: Problem Set 4 Assigned: 31/08/16, Due: 07/09/16 EE539R: Problem Set 4 Assigned: 3/08/6, Due: 07/09/6. Cover and Thomas: Problem 3.5 Sets defined by probabilities: Define the set C n (t = {x n : P X n(x n 2 nt } (a We have = P X n(x n P X n(x n 2 nt

More information

Aberrant Behavior Detection in Time Series for Monitoring Business-Critical Metrics (DRAFT)

Aberrant Behavior Detection in Time Series for Monitoring Business-Critical Metrics (DRAFT) Aberrant Behavior Detection in Time Series for Monitoring Business-Critical Metrics (DRAFT) Evan Miller IMVU, Inc. emiller@imvu.com Oct. 28, 2007 1 Abstract Detecting failures swiftly is a key process

More information

Clustering with k-means and Gaussian mixture distributions

Clustering with k-means and Gaussian mixture distributions Clustering with k-means and Gaussian mixture distributions Machine Learning and Category Representation 2014-2015 Jakob Verbeek, ovember 21, 2014 Course website: http://lear.inrialpes.fr/~verbeek/mlcr.14.15

More information

Fundamentals. CS 281A: Statistical Learning Theory. Yangqing Jia. August, Based on tutorial slides by Lester Mackey and Ariel Kleiner

Fundamentals. CS 281A: Statistical Learning Theory. Yangqing Jia. August, Based on tutorial slides by Lester Mackey and Ariel Kleiner Fundamentals CS 281A: Statistical Learning Theory Yangqing Jia Based on tutorial slides by Lester Mackey and Ariel Kleiner August, 2011 Outline 1 Probability 2 Statistics 3 Linear Algebra 4 Optimization

More information

11 - The linear model

11 - The linear model 11-1 The linear model S. Lall, Stanford 2011.02.15.01 11 - The linear model The linear model The joint pdf and covariance Example: uniform pdfs The importance of the prior Linear measurements with Gaussian

More information

CIFAR Lectures: Non-Gaussian statistics and natural images

CIFAR Lectures: Non-Gaussian statistics and natural images CIFAR Lectures: Non-Gaussian statistics and natural images Dept of Computer Science University of Helsinki, Finland Outline Part I: Theory of ICA Definition and difference to PCA Importance of non-gaussianity

More information

Chapter 2. Discrete Distributions

Chapter 2. Discrete Distributions Chapter. Discrete Distributions Objectives ˆ Basic Concepts & Epectations ˆ Binomial, Poisson, Geometric, Negative Binomial, and Hypergeometric Distributions ˆ Introduction to the Maimum Likelihood Estimation

More information

Lecture 7 Spectral methods

Lecture 7 Spectral methods CSE 291: Unsupervised learning Spring 2008 Lecture 7 Spectral methods 7.1 Linear algebra review 7.1.1 Eigenvalues and eigenvectors Definition 1. A d d matrix M has eigenvalue λ if there is a d-dimensional

More information

Math 416 Lecture 2 DEFINITION. Here are the multivariate versions: X, Y, Z iff P(X = x, Y = y, Z =z) = p(x, y, z) of X, Y, Z iff for all sets A, B, C,

Math 416 Lecture 2 DEFINITION. Here are the multivariate versions: X, Y, Z iff P(X = x, Y = y, Z =z) = p(x, y, z) of X, Y, Z iff for all sets A, B, C, Math 416 Lecture 2 DEFINITION. Here are the multivariate versions: PMF case: p(x, y, z) is the joint Probability Mass Function of X, Y, Z iff P(X = x, Y = y, Z =z) = p(x, y, z) PDF case: f(x, y, z) is

More information

Ph.D. Qualifying Exam Friday Saturday, January 3 4, 2014

Ph.D. Qualifying Exam Friday Saturday, January 3 4, 2014 Ph.D. Qualifying Exam Friday Saturday, January 3 4, 2014 Put your solution to each problem on a separate sheet of paper. Problem 1. (5166) Assume that two random samples {x i } and {y i } are independently

More information

Mark your answers ON THE EXAM ITSELF. If you are not sure of your answer you may wish to provide a brief explanation.

Mark your answers ON THE EXAM ITSELF. If you are not sure of your answer you may wish to provide a brief explanation. CS 189 Spring 2015 Introduction to Machine Learning Midterm You have 80 minutes for the exam. The exam is closed book, closed notes except your one-page crib sheet. No calculators or electronic items.

More information

Linear Algebra Notes. Lecture Notes, University of Toronto, Fall 2016

Linear Algebra Notes. Lecture Notes, University of Toronto, Fall 2016 Linear Algebra Notes Lecture Notes, University of Toronto, Fall 2016 (Ctd ) 11 Isomorphisms 1 Linear maps Definition 11 An invertible linear map T : V W is called a linear isomorphism from V to W Etymology:

More information

Statistics. Lecture 2 August 7, 2000 Frank Porter Caltech. The Fundamentals; Point Estimation. Maximum Likelihood, Least Squares and All That

Statistics. Lecture 2 August 7, 2000 Frank Porter Caltech. The Fundamentals; Point Estimation. Maximum Likelihood, Least Squares and All That Statistics Lecture 2 August 7, 2000 Frank Porter Caltech The plan for these lectures: The Fundamentals; Point Estimation Maximum Likelihood, Least Squares and All That What is a Confidence Interval? Interval

More information

Utility Maximizing Routing to Data Centers

Utility Maximizing Routing to Data Centers 0-0 Utility Maximizing Routing to Data Centers M. Sarwat, J. Shin and S. Kapoor (Presented by J. Shin) Sep 26, 2011 Sep 26, 2011 1 Outline 1. Problem Definition - Data Center Allocation 2. How to construct

More information

10-725/36-725: Convex Optimization Prerequisite Topics

10-725/36-725: Convex Optimization Prerequisite Topics 10-725/36-725: Convex Optimization Prerequisite Topics February 3, 2015 This is meant to be a brief, informal refresher of some topics that will form building blocks in this course. The content of the

More information

SYSM 6303: Quantitative Introduction to Risk and Uncertainty in Business Lecture 4: Fitting Data to Distributions

SYSM 6303: Quantitative Introduction to Risk and Uncertainty in Business Lecture 4: Fitting Data to Distributions SYSM 6303: Quantitative Introduction to Risk and Uncertainty in Business Lecture 4: Fitting Data to Distributions M. Vidyasagar Cecil & Ida Green Chair The University of Texas at Dallas Email: M.Vidyasagar@utdallas.edu

More information

Least Squares Optimization

Least Squares Optimization Least Squares Optimization The following is a brief review of least squares optimization and constrained optimization techniques. I assume the reader is familiar with basic linear algebra, including the

More information

SAMPLING AND INVERSION

SAMPLING AND INVERSION SAMPLING AND INVERSION Darryl Veitch dveitch@unimelb.edu.au CUBIN, Department of Electrical & Electronic Engineering University of Melbourne Workshop on Sampling the Internet, Paris 2005 A TALK WITH TWO

More information

Introduction to Maximum Likelihood Estimation

Introduction to Maximum Likelihood Estimation Introduction to Maximum Likelihood Estimation Eric Zivot July 26, 2012 The Likelihood Function Let 1 be an iid sample with pdf ( ; ) where is a ( 1) vector of parameters that characterize ( ; ) Example:

More information

ORF 522. Linear Programming and Convex Analysis

ORF 522. Linear Programming and Convex Analysis ORF 522 Linear Programming and Convex Analysis The Simplex Method Marco Cuturi Princeton ORF-522 1 Reminder: Basic Feasible Solutions, Extreme points, Optima Some important theorems last time for standard

More information

Lecture 7: Simulation of Markov Processes. Pasi Lassila Department of Communications and Networking

Lecture 7: Simulation of Markov Processes. Pasi Lassila Department of Communications and Networking Lecture 7: Simulation of Markov Processes Pasi Lassila Department of Communications and Networking Contents Markov processes theory recap Elementary queuing models for data networks Simulation of Markov

More information

Lecture 11: Quantum Information III - Source Coding

Lecture 11: Quantum Information III - Source Coding CSCI5370 Quantum Computing November 25, 203 Lecture : Quantum Information III - Source Coding Lecturer: Shengyu Zhang Scribe: Hing Yin Tsang. Holevo s bound Suppose Alice has an information source X that

More information

Queueing Theory and Simulation. Introduction

Queueing Theory and Simulation. Introduction Queueing Theory and Simulation Based on the slides of Dr. Dharma P. Agrawal, University of Cincinnati and Dr. Hiroyuki Ohsaki Graduate School of Information Science & Technology, Osaka University, Japan

More information

Lecture 13 Fundamentals of Bayesian Inference

Lecture 13 Fundamentals of Bayesian Inference Lecture 13 Fundamentals of Bayesian Inference Dennis Sun Stats 253 August 11, 2014 Outline of Lecture 1 Bayesian Models 2 Modeling Correlations Using Bayes 3 The Universal Algorithm 4 BUGS 5 Wrapping Up

More information

Privacy in Statistical Databases

Privacy in Statistical Databases Privacy in Statistical Databases Individuals x 1 x 2 x n Server/agency ) answers. A queries Users Government, researchers, businesses or) Malicious adversary What information can be released? Two conflicting

More information

LECTURE 2 LINEAR REGRESSION MODEL AND OLS

LECTURE 2 LINEAR REGRESSION MODEL AND OLS SEPTEMBER 29, 2014 LECTURE 2 LINEAR REGRESSION MODEL AND OLS Definitions A common question in econometrics is to study the effect of one group of variables X i, usually called the regressors, on another

More information

Business Statistics 41000: Homework # 2 Solutions

Business Statistics 41000: Homework # 2 Solutions Business Statistics 4000: Homework # 2 Solutions Drew Creal February 9, 204 Question #. Discrete Random Variables and Their Distributions (a) The probabilities have to sum to, which means that 0. + 0.3

More information

On Tandem Blocking Queues with a Common Retrial Queue

On Tandem Blocking Queues with a Common Retrial Queue On Tandem Blocking Queues with a Common Retrial Queue K. Avrachenkov U. Yechiali Abstract We consider systems of tandem blocking queues having a common retrial queue. The model represents dynamics of short

More information

Inderjit Dhillon The University of Texas at Austin

Inderjit Dhillon The University of Texas at Austin Inderjit Dhillon The University of Texas at Austin ( Universidad Carlos III de Madrid; 15 th June, 2012) (Based on joint work with J. Brickell, S. Sra, J. Tropp) Introduction 2 / 29 Notion of distance

More information

DS-GA 1002 Lecture notes 0 Fall Linear Algebra. These notes provide a review of basic concepts in linear algebra.

DS-GA 1002 Lecture notes 0 Fall Linear Algebra. These notes provide a review of basic concepts in linear algebra. DS-GA 1002 Lecture notes 0 Fall 2016 Linear Algebra These notes provide a review of basic concepts in linear algebra. 1 Vector spaces You are no doubt familiar with vectors in R 2 or R 3, i.e. [ ] 1.1

More information