Basic concepts of probability theory

Similar documents
Basic concepts of probability theory

Basic concepts of probability theory

Queuing Theory and Stochas St t ochas ic Service Syste y ms Li Xia

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

CDA6530: Performance Models of Computers and Networks. Chapter 2: Review of Practical Random Variables

Bulk input queue M [X] /M/1 Bulk service queue M/M [Y] /1 Erlangian queue M/E k /1

CDA5530: Performance Models of Computers and Networks. Chapter 2: Review of Practical Random Variables

Slides 8: Statistical Models in Simulation

CDA6530: Performance Models of Computers and Networks. Chapter 2: Review of Practical Random

1 Review of Probability

Lecture Notes 1 Probability and Random Variables. Conditional Probability and Independence. Functions of a Random Variable

Chapter 2: Random Variables

Chapter 5. Chapter 5 sections

6.1 Moment Generating and Characteristic Functions

1 Random Variable: Topics

System Simulation Part II: Mathematical and Statistical Models Chapter 5: Statistical Models

Lecture Notes 1 Probability and Random Variables. Conditional Probability and Independence. Functions of a Random Variable

A Probability Primer. A random walk down a probabilistic path leading to some stochastic thoughts on chance events and uncertain outcomes.

continuous random variables

Random variables. DS GA 1002 Probability and Statistics for Data Science.

1.1 Review of Probability Theory

1 Basic concepts from probability theory

CS145: Probability & Computing

Lecture 2: Repetition of probability theory and statistics

3. Probability and Statistics

Things to remember when learning probability distributions:

2 Random Variable Generation

General Random Variables

Part I Stochastic variables and Markov chains

Basics of Stochastic Modeling: Part II

Lecture 1: August 28

1 Inverse Transform Method and some alternative algorithms

n! (k 1)!(n k)! = F (X) U(0, 1). (x, y) = n(n 1) ( F (y) F (x) ) n 2

EE/CpE 345. Modeling and Simulation. Fall Class 5 September 30, 2002

STAT Chapter 5 Continuous Distributions

Chapter 2 Queueing Theory and Simulation

Continuous Probability Distributions. Uniform Distribution

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

Northwestern University Department of Electrical Engineering and Computer Science

Queueing Theory and Simulation. Introduction

Lecture Notes 7 Random Processes. Markov Processes Markov Chains. Random Processes

Brief Review of Probability

CSE 312, 2017 Winter, W.L. Ruzzo. 7. continuous random variables

Generation from simple discrete distributions

Probability Distributions Columns (a) through (d)

Chapter 5 continued. Chapter 5 sections

Continuous Random Variables and Continuous Distributions

Chapter 5. Statistical Models in Simulations 5.1. Prof. Dr. Mesut Güneş Ch. 5 Statistical Models in Simulations

Twelfth Problem Assignment

HW7 Solutions. f(x) = 0 otherwise. 0 otherwise. The density function looks like this: = 20 if x [10, 90) if x [90, 100]

Random Variables. Cumulative Distribution Function (CDF) Amappingthattransformstheeventstotherealline.

Lecturer: Olga Galinina

Stat 100a, Introduction to Probability.

3 Multiple Discrete Random Variables

Computer Science, Informatik 4 Communication and Distributed Systems. Simulation. Discrete-Event System Simulation. Dr.

5. Random Vectors. probabilities. characteristic function. cross correlation, cross covariance. Gaussian random vectors. functions of random vectors

1 Review of Probability and Distributions

Lecture 4: Random Variables and Distributions

STAT/MATH 395 A - PROBABILITY II UW Winter Quarter Moment functions. x r p X (x) (1) E[X r ] = x r f X (x) dx (2) (x E[X]) r p X (x) (3)

Probability Models. 4. What is the definition of the expectation of a discrete random variable?

4 Branching Processes

Probability distributions. Probability Distribution Functions. Probability distributions (contd.) Binomial distribution

It can be shown that if X 1 ;X 2 ;:::;X n are independent r.v. s with

Moments. Raw moment: February 25, 2014 Normalized / Standardized moment:

Math Spring Practice for the final Exam.

Algorithms for Uncertainty Quantification

Probability and Distributions

Recap. Probability, stochastic processes, Markov chains. ELEC-C7210 Modeling and analysis of communication networks

EEL 5544 Noise in Linear Systems Lecture 30. X (s) = E [ e sx] f X (x)e sx dx. Moments can be found from the Laplace transform as

EE4601 Communication Systems

Chapter 2. Random Variable. Define single random variables in terms of their PDF and CDF, and calculate moments such as the mean and variance.

EE 505 Introduction. What do we mean by random with respect to variables and signals?

Tom Salisbury

SOLUTION FOR HOMEWORK 12, STAT 4351

CHAPTER 6. 1, if n =1, 2p(1 p), if n =2, n (1 p) n 1 n p + p n 1 (1 p), if n =3, 4, 5,... var(d) = 4var(R) =4np(1 p).

STAT 430/510: Lecture 15

Lecture 3 Continuous Random Variable

Chapter 3: Random Variables 1

Discrete Distributions

Part IA Probability. Definitions. Based on lectures by R. Weber Notes taken by Dexter Chua. Lent 2015

Continuous Random Variables

Discrete Random Variables

IE 303 Discrete-Event Simulation

Stochastic Models in Computer Science A Tutorial

Closed book and notes. 60 minutes. Cover page and four pages of exam. No calculators.

Physics 403 Probability Distributions II: More Properties of PDFs and PMFs

MA/ST 810 Mathematical-Statistical Modeling and Analysis of Complex Systems

Preliminary Statistics. Lecture 3: Probability Models and Distributions

P (x). all other X j =x j. If X is a continuous random vector (see p.172), then the marginal distributions of X i are: f(x)dx 1 dx n

Distributions of Functions of Random Variables. 5.1 Functions of One Random Variable

Statistics, Data Analysis, and Simulation SS 2015

Probability and Statistics Concepts

Special Discrete RV s. Then X = the number of successes is a binomial RV. X ~ Bin(n,p).

Test Problems for Probability Theory ,

Multiple Random Variables

Random Variables. P(x) = P[X(e)] = P(e). (1)

ECE 302 Division 2 Exam 2 Solutions, 11/4/2009.

Guidelines for Solving Probability Problems

Random Variables. Saravanan Vijayakumaran Department of Electrical Engineering Indian Institute of Technology Bombay

ECE 313 Probability with Engineering Applications Fall 2000

Transcription:

Basic concepts of probability theory Random variable discrete/continuous random variable Transform Z transform, Laplace transform Distribution Geometric, mixed-geometric, Binomial, Poisson, exponential, Erlang, Hyper-exponential, phasetype, Markovian arrival process 1

Random variable X is denoted as random variable Discrete random variable, if X is discrete Continuous random variable, if X is continuous Distribution function, or called cdf, cumulative distribution function F(x)=Pr(X<x) Probability density/mass function f(x)=pr(x=x), x is discrete (probability mass function, pmf) f(x)= F(x)/ x, x is continuous (pdf) 2

Random variable Mean: E(X) Variance: Var(X), or σ 2 (X) σ 2 (X) = E{(X-E(X)) 2 }=E(X 2 )-E 2 (X) Standard deviation: σ(x) Covariance of two random variables X, Y Cov(X,Y)=E{(X-E(X))(Y-E(Y))} Correlation coefficient of X, Y r(x,y)=cov(x,y)/ σ(x)σ(y) -1 r(x,y) 1 3

Coefficient of variation: Coefficient of variation: c X c X = σ(x)/e(x) c X =0: deterministic c X <1: smooth c X =1: pure random c X >1: bursty more than variance Figures as example (interarrival time) t t t 4

Discrete Random Variables Probability mass function(pmf), P(n) Pn ( ) = Pr[ X= n] where 0 Pn ( ) 1, Pn ( ) = 1 Cumulative distribution function(cdf), F(x) Fx ( ) = Pr[ X x] = Pn ( ) where F(0) = 0 F( ) = 1 n x Fb ( ) Fa ( ), if b a Fb ( ) Fa ( ) = Pr[ a< X< b] n Suppose the random variable X is positive 5

Discrete Random Variables Mean and Variance Mean: x = np(n) n Variance: 2 2 σ = ( n x) P(n) n 6

Continuous Random Variables Probability density function(pdf), f(x) f( x) 0 Pr[ a X b] = f ( x) dx 0 f ( x) dx = 1 b a Cumulative distribution function(cdf), F(x) F( x) = Pr[ X x] = f ( y) dy df( x) f( x) = dx 0 x 7

Continuous Random Variables Mean and Variance Mean: x = xf ( x) dx = xdf( x) 0 0 x = (1 F ( x )) dx 0 Variance: σ 2 2 2 2 = = ( x x) f ( x) dx x f ( x) dx x 0 0 8

Z-transform for discrete distribution Z-transform is also called generating function P(z): Z-transform of discrete r.v. X, p(n)=pr(x=n), assume n=0,1,2, Pz ( ) = Ez [ X ] = pnz ( ) n= 0 Property P(0) = p(0), P(1) = 1, P (1) = EX ( ) P (1) =? n 9

Laplace-transform for continuous distribution F*(s): Laplace transform of a continuous r.v. X pdf of X is f(x), cdf of X is F(x), assume x 0 * sx sx sx F s E e = = e f x dx = e df x 0 0 ( ) [ ] ( ) ( ) Property Shortcut to calculate the k-moment of X * * *( k) k k F(0) = 1, F (0) = EX ( ), F (0) = ( 1) EX ( ) * sx F () s = s e F() x dx 0 10

Geometric distribution p: success probability in a Bernoulli trial X: the number of Bernoulli trials needed to first get one success g(k;p)=pr{x=k}=p(1-p) k-1, k=1,2, Property E(X)=1/p Var(X)=(1-p)/p 2 c X =(1-p) ½ Curve of probability mass function Discrete version of exponential distribution 11

Mixed Geometric distribution n set of independent Bernoulli trials, with success probability p i, respectively, i=1,2,,n Mixed probability is θ i, i=1,2,,n X: the number of Bernoulli trials first get one success of any one from the n set trials n k pmf: gb( k; θ, p) = θipi(1 pi) EX ( ) = θi / n i= 1 p i i= 1 Discrete version of hyper-exponential distribution 12

Binomial distribution p: success probability in a Bernoulli trial X: the number of successes during n Bernoulli trials Probability function n r Prnp ( ;, ) = Pr[ X= r] = p(1 p) r Property E(X)=np n r 13

Negative binomial distribution Definition (also called Pascal distribution) Bernoulli trial with success prob. p, predefined number k of failures has occurred, stop the random number of successes, X, obeys NB distr. k+ r 1 r nb( rk ;, p) = Pr[ X= r] = p(1 p) r Not required, for your reference k 14

Poisson distribution A Poisson random variable X with parameter λ has probability distribution n λ λ PX ( = n) = e, n= 0,1, 2,... n! For the Poisson distribution, it holds X = σ ( X) = λ, cx = 2 1 λ 15

Exponential Distribution (with Parameter μ) Pdf and cdf are f( x) = µ e µ x F( x) = 1 e, x 0 1 2 1 σ x x = σ x = c 1 2 x = = µ µ x f(x) µ 1 Pure random F(x) 0 x 0 x 16

Exponential distribution If X 1, X 2,, X n are independent exponential random variables with parameter μ 1, μ 2,, μ n Then, Y = min(x 1, X 2,, X n ) is an exponential random variable with parameter μ = μ 1 + μ 2 + + μ n How about Z = max(x 1, X 2,, X n )? Prove it as homework 17

Exponential distribution has memoryless property PX ( > x+ t) PX ( > x+ t X> t) = PX ( > t) µ ( x+ t) e µ x = = e = PX ( > x) µ t e P( t < X < x+ t X > t) = P( X < x) = F( x) = 1 e µx Memoryless: Future state only depends on the current state, independent of the history 18

Erlang distribution X: Erlang-k random variable, X= X 1 +X 2 + +X k, X i are independent random variables obeying exponential distribution with para. μ Denote this distribution as E k (μ), its pdf is Its cdf is 19

Erlang distribution Mean, variance, squared coefficient of variation Model smooth data traffic barber shop pdf curve is 20

Hyperexponential distribution X is with prob. α i to select exponential r.v. X i with para. μ i, i=1,,k, denote this distribution as H k k it Its pdf is f( t) = αµ i ie µ, t 0 Its mean is Its variance is i= 1 k αi EX ( ) = µ Its coefficient of variation 1 i= 1 Model bursty data traffic i k k 2 α i α i σ ( X ) = 2 2 i= 1 µ i i= 1 µ i 2 α 1 α k µ 1 1 µ k k 21

Laplace distribution also called the double exponential distribution two exponential distributions (with an additional location parameter) spliced together back-to-back Pdf of Laplace(μμ, bb): 1 x µ f( x µ, b) = exp 2b b μμ is location parameter, b is scale parameter 22

Phase-type distribution (discrete PH type distribution time case) the first passage time to the absorbing state of a discrete time Markov chain Two para., T: part of transition probability matrix, α: the initial state distribution, n phase k cdf:, pmf: Example Fk ( ) = 1 αte k 1 f( k) = αt t 0.3 0.4 0.3 T t P = 0.2 0.6 0.2 = 0 1 0 0 1 α = [0.3,0.7] Te+ t = e 23

Phase-type distribution (discrete time case) Example Geometric distribution, 1 phase 1 p p T t P = = 0 1 α = [1] 0 1 pmf: Mix-Geometric distribution, n phase 0.3 0 0.7 T t P = 0 0.8 0.2 = 0 1 0 0 1 pmf: f( k) = T t = p(1 p) k 1 k 1 α α = [0.5,0.5] f( k) = T t = 0.5 0.7 0.3 + 0.5 0.2 0.8 k 1 k 1 k 1 α 24

PH type distribution (continuous PH type distribution time case) the first passage time to the absorbing state of a continuous time Markov process Two para., T: part of the transition rate matrix, α: the initial state distribution, n phase cdf: Example Tx F( x) = 1 αe e 5 4 1 T t B = 2 6 4 = 0 0 0 0 0 -pdf: α = [0.8,0.2] Te+ t = 0 ff xx = αα ee TTTT tt 25

PH type distribution (continuous Example time case) Exponential distribution: F( x) 1 e Tx = α e= 1 e λ Erlang distribution with (m,λ): α = [1,0,0,...,0] Hyper-exponential distribution α = [ θ1, θ2,..., θ n ] x α = 1, T = λ T t B = 0 0 λ λ λ λ T =.... λ λ1 λ 2 T =... λn 26

Coxian distribution A special case of PH type distribution µ 1 p1µ 1 0 0... (1 p1) µ 1 0 µ 2 p2µ 2 0... (1 p2) µ 2 B = 0 0... µ k 1 pk 1µ k 1 (1 pk 1) µ k 1 0 0... 0 µ k µ k 0 0... 0 0 0 α = [1,0,0,...,0] k phase 27

Pareto distribution Pareto(α): a distribution with a power-law tail pdf: cdf: f x x x f x Usually, 0<α<2 α 1 ( ) = α, 1; otherwise, ( ) = 0 α F( x) = 1 x, x 1; otherwise, F( x) = 0 Pareto distribution decays much more slowly than exponential distribution We call it has heavy tail or fat tail Heavy tail is very important for practical data traffic 28

Relation to Central Limit Theorem CLT: X i are i.i.d. r.v. s with mean μ and variance σ 2, define Z n =(X 1 + + X n - nμ)/(σ n), then Z n converges to the standard normal distribution CLT can explain many natural phenomena obey normal distribution Binomial(n,p) is a sum of i.i.d. Bernoulli(p) r.v. s, it converges to a normal distribution when n is large Poisson(λ) can be approximated by a normal distribution with mean λ and variance λ: sum of λ r.v. s with rate 1 Binomial and Poisson can be viewed as an approximation of discrete version of normal distribution, plot their pdf 29

Relation to Central Limit Theorem Binomial Distribution sum of n Geometric distribution r.v. s with probability p Poisson Distribution sum of λ Poisson distribution r.v. s with rate 1 30

An explanation to Additive White Gaussian Noise (AWGN) AWGN explanation Transmit a signal in a channel 100 i.i.d. sources make low noise, uniformly distributed among [-1, 1] Noise is additive, if the total amount of noise level is greater than 10 or less than -10, then it corrupts the signal; otherwise, no problem Calculate the probability of signal no corruption Calculation X~Unif(-1,1), E(X)=0, Var(X)=1/3; S 100 ~Norm(0,100/3) P(-10<S 100 <10)=Φ( 3)- Φ(- 3)=2Φ( 3)-1=2(0.9572)-1 =0.9144 31

A map of the relation of all these distributions Chalk drawing PH type distribution can almost model any distribution Convert stochastic process to Markovian model Widely used in queuing system, reliable system, regenerative system, etc. Discrete version v.s. continuous version Geometric v.s. Exponential Mixed Geo v.s. HyerExp Binomial, Poisson v.s. Normal Erlang (continuous r.v.) Normal(continuous r.v.) 32