SYSTEMS Identification Ali Karimpour Assistant Professor Ferdowsi University of Mashhad Reference: “System Identification Theory For The User” Lennart.

Slides:



Advertisements
Similar presentations
Properties of Least Squares Regression Coefficients
Advertisements

ECE 8443 – Pattern Recognition ECE 8423 – Adaptive Signal Processing Objectives: The Linear Prediction Model The Autocorrelation Method Levinson and Durbin.
CHAPTER 8 More About Estimation. 8.1 Bayesian Estimation In this chapter we introduce the concepts related to estimation and begin this by considering.
ECE 8443 – Pattern Recognition LECTURE 05: MAXIMUM LIKELIHOOD ESTIMATION Objectives: Discrete Features Maximum Likelihood Resources: D.H.S: Chapter 3 (Part.
1 12. Principles of Parameter Estimation The purpose of this lecture is to illustrate the usefulness of the various concepts introduced and studied in.
Lecture 3 Today: Statistical Review cont’d:
The General Linear Model. The Simple Linear Model Linear Regression.
Lecture 2 Today: Statistical Review cont’d:
AGC DSP AGC DSP Professor A G Constantinides©1 A Prediction Problem Problem: Given a sample set of a stationary processes to predict the value of the process.
Visual Recognition Tutorial
The Simple Linear Regression Model: Specification and Estimation
280 SYSTEM IDENTIFICATION The System Identification Problem is to estimate a model of a system based on input-output data. Basic Configuration continuous.
SYSTEMS Identification Ali Karimpour Assistant Professor Ferdowsi University of Mashhad Reference: “System Identification Theory For The User” Lennart.
Ch 7.9: Nonhomogeneous Linear Systems
SYSTEMS Identification
SYSTEMS Identification Ali Karimpour Assistant Professor Ferdowsi University of Mashhad Reference: “System Identification Theory For The User” Lennart.
SYSTEMS Identification
SUMS OF RANDOM VARIABLES Changfei Chen. Sums of Random Variables Let be a sequence of random variables, and let be their sum:
Chapter 4 Multiple Regression.
SYSTEMS Identification Ali Karimpour Assistant Professor Ferdowsi University of Mashhad Reference: “System Identification Theory For The User” Lennart.
SYSTEMS Identification
SYSTEMS Identification
Multivariable Control Systems
Development of Empirical Models From Process Data
1 EE 616 Computer Aided Analysis of Electronic Networks Lecture 9 Instructor: Dr. J. A. Starzyk, Professor School of EECS Ohio University Athens, OH,
SYSTEMS Identification Ali Karimpour Assistant Professor Ferdowsi University of Mashhad Reference: “System Identification Theory For The User” Lennart.
Probability theory 2010 Conditional distributions  Conditional probability:  Conditional probability mass function: Discrete case  Conditional probability.
Topic4 Ordinary Least Squares. Suppose that X is a non-random variable Y is a random variable that is affected by X in a linear fashion and by the random.
Linear and generalised linear models Purpose of linear models Least-squares solution for linear models Analysis of diagnostics Exponential family and generalised.
Maximum likelihood (ML)
Christopher Dougherty EC220 - Introduction to econometrics (chapter 3) Slideshow: prediction Original citation: Dougherty, C. (2012) EC220 - Introduction.
1 10. Joint Moments and Joint Characteristic Functions Following section 6, in this section we shall introduce various parameters to compactly represent.
Relationships Among Variables
1 PREDICTION In the previous sequence, we saw how to predict the price of a good or asset given the composition of its characteristics. In this sequence,
Adaptive Signal Processing
Physics 114: Lecture 15 Probability Tests & Linear Fitting Dale E. Gary NJIT Physics Department.
Ch 8.1 Numerical Methods: The Euler or Tangent Line Method
CORRELATION & REGRESSION
ECE 8443 – Pattern Recognition LECTURE 06: MAXIMUM LIKELIHOOD AND BAYESIAN ESTIMATION Objectives: Bias in ML Estimates Bayesian Estimation Example Resources:
1 As we have seen in section 4 conditional probability density functions are useful to update the information about an event based on the knowledge about.
Model Inference and Averaging
Probabilistic and Statistical Techniques 1 Lecture 24 Eng. Ismail Zakaria El Daour 2010.
SYSTEMS Identification Ali Karimpour Assistant Professor Ferdowsi University of Mashhad Reference: “System Identification Theory For The User” Lennart.
ECE 8443 – Pattern Recognition ECE 8423 – Adaptive Signal Processing Objectives: Deterministic vs. Random Maximum A Posteriori Maximum Likelihood Minimum.
ECE 8443 – Pattern Recognition LECTURE 07: MAXIMUM LIKELIHOOD AND BAYESIAN ESTIMATION Objectives: Class-Conditional Density The Multivariate Case General.
CHAPTER 4 Adaptive Tapped-delay-line Filters Using the Least Squares Adaptive Filtering.
SUPA Advanced Data Analysis Course, Jan 6th – 7th 2009 Advanced Data Analysis for the Physical Sciences Dr Martin Hendry Dept of Physics and Astronomy.
Vector Norms and the related Matrix Norms. Properties of a Vector Norm: Euclidean Vector Norm: Riemannian metric:
Computational Intelligence: Methods and Applications Lecture 23 Logistic discrimination and support vectors Włodzisław Duch Dept. of Informatics, UMK Google:
PROBABILITY AND STATISTICS FOR ENGINEERING Hossein Sameti Department of Computer Engineering Sharif University of Technology Principles of Parameter Estimation.
SYSTEMS Identification Ali Karimpour Assistant Professor Ferdowsi University of Mashhad Reference: “System Identification Theory For The User” Lennart.
ECE 8443 – Pattern Recognition ECE 8527 – Introduction to Machine Learning and Pattern Recognition LECTURE 07: BAYESIAN ESTIMATION (Cont.) Objectives:
SYSTEMS Identification Ali Karimpour Assistant Professor Ferdowsi University of Mashhad Reference: “System Identification Theory For The User” Lennart.
An Introduction to Kalman Filtering by Arthur Pece
Estimation Method of Moments (MM) Methods of Moment estimation is a general method where equations for estimating parameters are found by equating population.
Joint Moments and Joint Characteristic Functions.
SYSTEMS Identification Ali Karimpour Assistant Professor Ferdowsi University of Mashhad Reference: “System Identification Theory.
SYSTEMS Identification Ali Karimpour Assistant Professor Ferdowsi University of Mashhad.
R. Kass/W03 P416 Lecture 5 l Suppose we are trying to measure the true value of some quantity (x T ). u We make repeated measurements of this quantity.
Computacion Inteligente Least-Square Methods for System Identification.
Statistics 350 Lecture 2. Today Last Day: Section Today: Section 1.6 Homework #1: Chapter 1 Problems (page 33-38): 2, 5, 6, 7, 22, 26, 33, 34,
Chapter 7. Classification and Prediction
Tutorial 9: Further Topics on Random Variables 2
Undergraduated Econometrics
SYSTEMS Identification
The Simple Linear Regression Model: Specification and Estimation
Further Topics on Random Variables: Derived Distributions
16. Mean Square Estimation
Further Topics on Random Variables: Derived Distributions
Further Topics on Random Variables: Derived Distributions
Presentation transcript:

SYSTEMS Identification Ali Karimpour Assistant Professor Ferdowsi University of Mashhad Reference: “System Identification Theory For The User” Lennart Ljung

lecture 9 Ali Karimpour Nov 2010 Lecture 9 Topics to be covered include: v Central Limit Theorem v The Prediction-Error approach: Basic theorem v Expression for the Asymptotic Variance v Frequency-Domain Expressions for the Asymptotic Variance v Distribution of Estimation for the correlation Approach v Distribution of Estimation for the Instrumental Variable Methods 2 Asymptotic Distribution of Parameter Estimators

lecture 9 Ali Karimpour Nov 2010 Overview 3 If convergence is guaranteed, then But, how fast does the estimate approach the limit? What is the probability distribution of ? The variance analysis of this chapter will reveal: a) The estimate converges to at a rate proportional to b) Distribution converges to a Gaussian distribution: N(0,Q) c) Covariance matrix Q, depends on - The number of samples/data set size: N, - The parameter sensitivity of the predictor: - The noise variance

lecture 9 Ali Karimpour Nov 2010 Overview 4 If convergence is guaranteed, then

lecture 9 Ali Karimpour Nov 2010 Central Limit Theorem Topics to be covered include: v Central Limit Theorem v The Prediction-Error approach: Basic Theorem v Expression for the Asymptotic Variance v Frequency-Domain Expressions for the Asymptotic Variance v Distribution of Estimation for the correlation Approach v Distribution of Estimation for the Instrumental Variable Methods 5

lecture 9 Ali Karimpour Nov Central Limit Theorem The mathematical tool needed for asymptotic variance analysis is “Central Limit” theorems. Example: Consider two independent random variable, X and Y, with the same uniform distribution, shown in Figure below. Define another random variable Z as the sum of X and Y: Z=X+Y. we can obtain the distribution of Z, as :

lecture 9 Ali Karimpour Nov In general, the PDF of a random variable approaches a Gaussian distribution, regardless of the PDF of each, as N gets larger. Central Limit Theorem Further, consider W=X+Y+Z. The resultant PDF is getting close to a Gaussian distribution The resultant PDF is getting close to a Gaussian distribution.

lecture 9 Ali Karimpour Nov Central Limit Theorem Let be a d-dimensional random variable with : Mean Cov Consider the sum of given by: Then, as N tends to infinity, the distribution of converges to the Gaussian distribution given by PDF:

lecture 9 Ali Karimpour Nov 2010 The Prediction-Error approach: Basic Theorem Topics to be covered include: v Central Limit Theorem v The Prediction-Error approach: Basic Theorem v Expression for the Asymptotic Variance v Frequency-Domain Expressions for the Asymptotic Variance v Distribution of Estimation for the correlation Approach v Distribution of Estimation for the Instrumental Variable Methods 9

lecture 9 Ali Karimpour Nov 2010 The Prediction-Error Approach 10 Then, with prime denoting differentiation with respect to, Expanding around gives: is a vector “between” Applying the Central Limit Theorem, we can obtain the distribution of estimate as N tends to infinity. Let be an estimate based on the prediction error method

lecture 9 Ali Karimpour Nov 2010 The Prediction-Error Approach 11 Assume that is nonsingular, then: Where as usual: To obtain the distribution of, and must be computed as N tends to infinity.

lecture 9 Ali Karimpour Nov 2010 The Prediction-Error Approach 12 For simplicity, we first assume that the predictor is given by a linear regression: The actual data is generated by ( is the parameter vector of the true system) So: Therefore:

lecture 9 Ali Karimpour Nov Let us treat as a random variable. Its mean is zero, since: The covariance is Consider: Appling the central limit Theorem: The Prediction-Error Approach

lecture 9 Ali Karimpour Nov 2010 The Prediction-Error Approach 14 Next, compute : And: Exercise1: Proof (I)

lecture 9 Ali Karimpour Nov 2010 The Prediction-Error Approach 15 We obtain: So:

lecture 9 Ali Karimpour Nov 2010 The Prediction-Error Approach 16 Theorem Consider the estimate determined by: and also we have: and also: Then Assume that the model structure is linear and uniformly stable and that the data set is subject to D1. Assume also that for a unique value interior to D M we have: where

lecture 9 Ali Karimpour Nov 2010 The Prediction-Error Approach 17 As stated formally in pervious Theorem, the distribution of converges to a Gaussian distribution for the broad class of system identification problems. This is called the asymptotic covariance matrix and it depends on (c) Noise variance This implies that the covariance of asymptotically converges to: (a) the number of samples/data set size: N (b) the parameter sensitivity of the predictor:

lecture 9 Ali Karimpour Nov 2010 Expression for the Asymptotic Variance Topics to be covered include: v Central Limit Theorem v The Prediction-Error approach: Basic Theorem v Expression for the Asymptotic Variance v Frequency-Domain Expressions for the Asymptotic Variance v Distribution of Estimation for the correlation Approach v Distribution of Estimation for the Instrumental Variable Methods 18

lecture 9 Ali Karimpour Nov Let us compute the covariance once again for the general case: Unlike the linear regression, the sensitivity is a function of θ, Expression for the Asymptotic Variance

lecture 9 Ali Karimpour Nov So Expression for the Asymptotic Variance Similarly Hence : Therefore The asymptotic variance is therefore a) inversely proportional to the number of samples, b) proportional to the noise variance, and c) Inversely related to the parameter sensitivity.

lecture 9 Ali Karimpour Nov A very important and useful aspect of expressions for the asymptotic covariance matrix is that it can be estimated from data. Having N data points and determined we may use: Since is not known, the asymptotic variance cannot be determined. sufficient data samples needed for assuming the model accuracy may be obtained. Expression for the Asymptotic Variance

lecture 9 Ali Karimpour Nov Consider the system Suppose that the coefficient for u(t-1) is known and the system is identified in the model structure and are two independent white noise with variances and respectively We have: Example : Covariance of LS Estimates Or Hence

lecture 9 Ali Karimpour Nov Hence To compute the covariance, square the first equation and take the expectation: Multiplying the first equation by y(t-1) and taking expectation gives: The last equality follows, since u(t) does not affect y(t) (due to the time delay ) Hence: Example : Covariance of LS Estimates

lecture 9 Ali Karimpour Nov Cov (a) = The actual value of a is considered as 0.1 for the previous example. Both e(t) and u(t) are considered white noise with variance 1 and number of data is 10. Estimated values for parameter a, for 100 independent experiment using LSE, is shown in the bellow Figure. Example : Covariance of LS Estimates

lecture 9 Ali Karimpour Nov Cov (a) = Example : Covariance of LS Estimates Now we increase the variance of u(t) from 1 to 10. But other parameters are the same. It can be seen that Cov(a) decreases by increasing input variance, as we expect.

lecture 9 Ali Karimpour Nov Now we increase the variance of e(t) from 1 to 10 and set the variance of u(t) in 1 and repeat the first experiment while a=0.1. It can be seen that Cov(a) increases by increasing noise variance, as we expect. Cov (a) = Example : Covariance of LS Estimates

lecture 9 Ali Karimpour Nov Example : Covariance of an MA(1) Parameter Consider the system is white noise with variance. The MA(1) model structure is used: Given the predictor 4.18: Differentiation w.r.t c gives At c=c 0 we have : If is the PEM estimate of c: Exercise2: Proof (I)

lecture 9 Ali Karimpour Nov For general Norm We have: Similarly: The expression for the asymptotic covariance matrix is rather complicated in general. Asymptotic Variance for general Norms.

lecture 9 Ali Karimpour Nov After straightforward calculations: The choice of in the criterion only acts as scaling of the covariance matrix Exercise3: Proof (I) where

lecture 9 Ali Karimpour Nov 2010 Frequency-Domain Expressions for the Asymptotic Variance Topics to be covered include: v Central Limit Theorem v The Prediction-Error approach: Basic Theorem v Expression for the Asymptotic Variance v Frequency-Domain Expressions for the Asymptotic Variance v Distribution of Estimation for the correlation Approach v Distribution of Estimation for the Instrumental Variable Methods 30

lecture 9 Ali Karimpour Nov Frequency-Domain Expressions for the Asymptotic Variance. The asymptotic variance has different expression in the frequency domain, which we will find useful for variance analysis and experiment design. Let transfer function and noise model be consolidated into a matrix The gradient of T, that is, the sensitivity of T to θ, is For a predictor, we have already defined W(q,θ,) and z(t), s.t.

lecture 9 Ali Karimpour Nov Therefore the predictor sensitivity is given by Where Substituting in the first equation: Frequency-Domain Expressions for the Asymptotic Variance.

lecture 9 Ali Karimpour Nov At (the true system), note and where Let be the spectrum matrix of : Using the familiar formula: Frequency-Domain Expressions for the Asymptotic Variance.

lecture 9 Ali Karimpour Nov For the noise spectrum, Using this in equation below: We have: The asymptotic variance in the frequency domain. Frequency-Domain Expressions for the Asymptotic Variance.

lecture 9 Ali Karimpour Nov 2010 Distribution of Estimation for the correlation Approach Topics to be covered include: v Central Limit Theorem v The Prediction-Error approach: Basic Theorem v Expression for the Asymptotic Variance v Frequency-Domain Expressions for the Asymptotic Variance v Distribution of Estimation for the correlation Approach v Distribution of Estimation for the Instrumental Variable Methods 35

lecture 9 Ali Karimpour Nov The Correlation Approach By Taylor expansion we have: We shall confine ourselves to the case study in Theorem 8.6, that is, and linearly generated instruments. We thus have:

lecture 9 Ali Karimpour Nov This is entirely analogous with the previous one obtained for PE approach, with he difference that in is replaced with in. The Correlation Approach

lecture 9 Ali Karimpour Nov The Correlation Approach Theorem : consider by Assume that is computed for a linear, uniformly stable model structure And that: is a uniformly stable family of filters. Assume also that that is nonsingular and that And the data set is subject to D1

lecture 9 Ali Karimpour Nov Under the assumption S M,there exists a value such that: For L(q)=1 Then The Correlation Approach

lecture 9 Ali Karimpour Nov The Correlation Approach

lecture 9 Ali Karimpour Nov Example 9.5 : Covarianc of Pseudolinear Regression Estimate for example 9.2 but suppose that the c estimate is determined by the PLR method that is: The Correlation Approach Here:

lecture 9 Ali Karimpour Nov The Correlation Approach

lecture 9 Ali Karimpour Nov The Correlation Approach

lecture 9 Ali Karimpour Nov 2010 Distribution of Estimation for the Instrumental Variable Methods Topics to be covered include: v Central Limit Theorem v The Prediction-Error approach: Basic Theorem v Expression for the Asymptotic Variance v Frequency-Domain Expressions for the Asymptotic Variance v Distribution of Estimation for the correlation Approach v Distribution of Estimation for the Instrumental Variable Methods 44

lecture 9 Ali Karimpour Nov Instrumental Variable Methods Suppose the true system is given as Where e(t) is white noise with variance independent of {u(t)}. then is independent of {u(t)} and hence of if the system operates in open loop. Thus is a solution to: We have:

lecture 9 Ali Karimpour Nov Instrumental Variable Methods To get an asymptotic distribution, we shall assume it is the only solution to. Introduce also the monic filter Intersecting into these Eqs.,

lecture 9 Ali Karimpour Nov Instrumental Variable Methods

lecture 9 Ali Karimpour Nov 2010 Example 9.6 Covariance of an IV estimate : Consider the system : 48 Model structure : And let a be estimated by the IV method using the instrument is: And L(q) = 1 Instrumental Variable Methods

lecture 9 Ali Karimpour Nov Instrumental Variable Methods