Advanced Topics in Optimization. Piecewise Linear Approximation of a Nonlinear Function

Similar documents
Fuzzy approach to solve multi-objective capacitated transportation problem

( ) 2 ( ) ( ) Problem Set 4 Suggested Solutions. Problem 1

Linear system of the Schrödinger equation Notes on Quantum Mechanics

Kernel Methods and SVMs Extension

A Quadratic Cumulative Production Model for the Material Balance of Abnormally-Pressured Gas Reservoirs F.E. Gonzalez M.S.

A Quadratic Cumulative Production Model for the Material Balance of Abnormally-Pressured Gas Reservoirs F.E. Gonzalez M.S.

Dr. Shalabh Department of Mathematics and Statistics Indian Institute of Technology Kanpur

Solutions to exam in SF1811 Optimization, Jan 14, 2015

APPENDIX A Some Linear Algebra

Lecture Notes on Linear Regression

Fuzzy Set Approach to Solve Multi-objective Linear plus Fractional Programming Problem

Module 2. Random Processes. Version 2 ECE IIT, Kharagpur

princeton univ. F 17 cos 521: Advanced Algorithm Design Lecture 7: LP Duality Lecturer: Matt Weinberg

On the Multicriteria Integer Network Flow Problem

8.1 Arc Length. What is the length of a curve? How can we approximate it? We could do it following the pattern we ve used before

Mean Field / Variational Approximations

Transfer Functions. Convenient representation of a linear, dynamic model. A transfer function (TF) relates one input and one output: ( ) system

Introduction to Algorithms

Modelli Clamfim Equazioni differenziali 7 ottobre 2013

Polynomial Regression Models

Using T.O.M to Estimate Parameter of distributions that have not Single Exponential Family

The Expectation-Maximization Algorithm

Modelli Clamfim Equazioni differenziali 22 settembre 2016

Assortment Optimization under MNL

Non-Ideality Through Fugacity and Activity

Module 3 LOSSY IMAGE COMPRESSION SYSTEMS. Version 2 ECE IIT, Kharagpur

Appendix for Causal Interaction in Factorial Experiments: Application to Conjoint Analysis

PARTIAL QUOTIENTS AND DISTRIBUTION OF SEQUENCES. Department of Mathematics University of California Riverside, CA

Lecture 10 Support Vector Machines II

Foundations of Arithmetic

Solving Nonlinear Differential Equations by a Neural Network Method

Digital PI Controller Equations

CHAPTER 5 NUMERICAL EVALUATION OF DYNAMIC RESPONSE

Departure Process from a M/M/m/ Queue

Lecture 10 Support Vector Machines. Oct

( ) [ ( k) ( k) ( x) ( ) ( ) ( ) [ ] ξ [ ] [ ] [ ] ( )( ) i ( ) ( )( ) 2! ( ) = ( ) 3 Interpolation. Polynomial Approximation.

Lectures - Week 4 Matrix norms, Conditioning, Vector Spaces, Linear Independence, Spanning sets and Basis, Null space and Range of a Matrix

Introduction to Algorithms

= = = (a) Use the MATLAB command rref to solve the system. (b) Let A be the coefficient matrix and B be the right-hand side of the system.

An Interactive Optimisation Tool for Allocation Problems

The Dirac Equation for a One-electron atom. In this section we will derive the Dirac equation for a one-electron atom.

LINEAR REGRESSION ANALYSIS. MODULE IX Lecture Multicollinearity

Report on Image warping

The Minimum Universal Cost Flow in an Infeasible Flow Network

Interactive Bi-Level Multi-Objective Integer. Non-linear Programming Problem

ECEN 667 Power System Stability Lecture 21: Modal Analysis

ANSWERS. Problem 1. and the moment generating function (mgf) by. defined for any real t. Use this to show that E( U) var( U)

Numerical Solution of Ordinary Differential Equations

Chapter 3 Describing Data Using Numerical Measures

Lecture 12: Discrete Laplacian

Limited Dependent Variables

Not-for-Publication Appendix to Optimal Asymptotic Least Aquares Estimation in a Singular Set-up

EEE 241: Linear Systems

xp(x µ) = 0 p(x = 0 µ) + 1 p(x = 1 µ) = µ

Mixture of Gaussians Expectation Maximization (EM) Part 2

Quantum Mechanics for Scientists and Engineers. David Miller

Convexity preserving interpolation by splines of arbitrary degree

Some modelling aspects for the Matlab implementation of MMA

Hidden Markov Model Cheat Sheet

Solutions for Tutorial 1

Yong Joon Ryang. 1. Introduction Consider the multicommodity transportation problem with convex quadratic cost function. 1 2 (x x0 ) T Q(x x 0 )

Topology optimization of plate structures subject to initial excitations for minimum dynamic performance index

LECTURE 9 CANONICAL CORRELATION ANALYSIS

MATH 829: Introduction to Data Mining and Analysis The EM algorithm (part 2)

Module 9. Lecture 6. Duality in Assignment Problems

Matching Dyadic Distributions to Channels

Conservative Surrogate Model using Weighted Kriging Variance for Sampling-based RBDO

Solutions HW #2. minimize. Ax = b. Give the dual problem, and make the implicit equality constraints explicit. Solution.

AN EXTENDED MPC CONVERGENCE CONDITION

Singular Value Decomposition: Theory and Applications

Integrals and Invariants of Euler-Lagrange Equations

Managing Capacity Through Reward Programs. on-line companion page. Byung-Do Kim Seoul National University College of Business Administration

U.C. Berkeley CS294: Beyond Worst-Case Analysis Luca Trevisan September 5, 2017

System in Weibull Distribution

Lecture 16 Statistical Analysis in Biomaterials Research (Part II)

Confidence intervals for weighted polynomial calibrations

Bézier curves. Michael S. Floater. September 10, These notes provide an introduction to Bézier curves. i=0

NUMERICAL DIFFERENTIATION

Inner Product. Euclidean Space. Orthonormal Basis. Orthogonal

Lecture 17 : Stochastic Processes II

Some Notes on Consumer Theory

e - c o m p a n i o n

NECESSARY AND SUFFICIENT CONDITIONS FOR ALMOST REGULARITY OF UNIFORM BIRKHOFF INTERPOLATION SCHEMES. by Nicolae Crainic

Hongyi Miao, College of Science, Nanjing Forestry University, Nanjing ,China. (Received 20 June 2013, accepted 11 March 2014) I)ϕ (k)

6.3.4 Modified Euler s method of integration

PID Controller Design Based on Second Order Model Approximation by Using Stability Boundary Locus Fitting

Lesson 16: Basic Control Modes

Global Optimization of Truss. Structure Design INFORMS J. N. Hooker. Tallys Yunes. Slide 1

Lecture 20: November 7

Predictive Control of a Boiler-turbine System

The internal structure of natural numbers and one method for the definition of large prime numbers

CS : Algorithms and Uncertainty Lecture 17 Date: October 26, 2016

Mechanics Physics 151

Generalized Linear Methods

The Second Anti-Mathima on Game Theory

MIMA Group. Chapter 2 Bayesian Decision Theory. School of Computer Science and Technology, Shandong University. Xin-Shun SDU

8.4 COMPLEX VECTOR SPACES AND INNER PRODUCTS

CIVL 8/7117 Chapter 10 - Isoparametric Formulation 42/56

Lecture 3 Examples and Problems

5 The Rational Canonical Form

Transcription:

Advanced Tocs n Otmzaton Pecewse Lnear Aroxmaton of a Nonlnear Functon Otmzaton Methods: M8L

Introducton and Objectves Introducton There exsts no general algorthm for nonlnear rogrammng due to ts rregular behavor Nonlnear roblems can be solved by frst reresentng the nonlnear functon (both objectve functon and constrants) by a set of lnear functons and then aly smlex method to solve ths usng some restrctons Objectves To dscuss the varous methods to aroxmate a nonlnear functon usng lnear functons To demonstrate ths usng a numercal examle Otmzaton Methods: M8L

Pecewse Lnearzaton A nonlnear sngle varable functon f(x) can be aroxmated by a ecewse lnear functon Geometrcally, f(x) can be shown as a curve beng reresented as a set of connected lne segments 3 Otmzaton Methods: M8L

Pecewse Lnearzaton: Method Consder an otmzaton functon havng only one nonlnear term f(x) Let the x-axs of the nonlnear functon f(x) be dvded by breakng onts t, t, t,, t Corresondng functon values be f(t ), f(t ),, f(t ) If x can take values n the nterval 0 x X, then the breakng onts can be shown as 0 t < t <... < t X 4 Otmzaton Methods: M8L

Pecewse Lnearzaton: Method contd. Exress x as a weghted average of these breakng onts x = w t + w t +... + Functon f(x) can be exressed as. e., x = w t w t 5 f where ( x) = w f ( t ) + w f ( t ) +... + w f ( t ) = w f ( t ) w = Otmzaton Methods: M8L

Pecewse Lnearzaton: Method contd. Fnally the model can be exressed as Max or Mn f ( x) = w f ( ) subject to the addtonal constrants w t w = x = t 6 Otmzaton Methods: M8L

Pecewse Lnearzaton: Method contd. Ths lnearly aroxmated model can be solved usng smlex method wth some restrctons Restrcted condton: There should not be more than two w n the bass and Two w can take ostve values only f they are adjacent..e., f x takes the value between t and t +, then only w and w + (contrbutng weghts to the value of x ) wll be ostve, rest all weghts be zero In general, for an objectve functon consstng of n varables ( n terms) reresented as ( x) = f ( x ) + f ( x ) +... ( ) Max or Mn f + f n x n 7 Otmzaton Methods: M8L

Pecewse Lnearzaton: Method contd. subjected to m constrants g ( x ) + g ( x ) +... + g ( x ) b for j,,..., m j j nj n j = 8 The lnear aroxmaton of ths roblem s Max or Mn subjected to n n k = k = w k w w k k g f k kj ( t ( t k k ) ) b j for j =,,..., m = for k =,,..., n Otmzaton Methods: M8L

Pecewse Lnearzaton: Method 9 x s exressed as a sum, nstead of exressng as the weghted sum of the breakng onts as n the revous method x = t + u + u +... + u where u s the ncrement of the varable x n the nterval.e., the bound of u s 0 u t+ The functon f(x) can be exressed as ( x ) = f ( t ) u + α f = where α reresents the sloe of the lnear aroxmaton between the onts t + and t f ( t+ ) f ( t ) α = t t = t + u + t ( t, t + ) Otmzaton Methods: M8L

Pecewse Lnearzaton: Method contd. Fnally the model can be exressed as Max or Mn f α ( x) = f ( t ) + subjected to addtonal constrants u t + u = 0 u t+ x t, =,,..., 0 Otmzaton Methods: M8L

Pecewse Lnearzaton: Numercal Examle The examle below llustrates the alcaton of method Consder the objectve functon Maxmze f = + 3 x x subject to x 0 x + x x 0 4 5 The roblem s already n searable form (.e., each term conssts of only one varable). Otmzaton Methods: M8L

Pecewse Lnearzaton: Numercal Examle contd. Slt u the objectve functon and constrant nto two arts where and are treated as lnear varables as they are n lnear form f g f = g = f g ( x ) + f( x ) ( x ) + g ( x ) 3 ( x ) = x ; f( x ) = x ( x ) = x ; g( x ) = x f ( ) ( ) x g x Otmzaton Methods: M8L

Pecewse Lnearzaton: Numercal Examle contd. Consder fve breakng onts for x f ( x ) can be wrtten as, f 5 ( x ) w f ( t ) = = w 0 + w + w3 8 + w4 7 + w5 64 3 Otmzaton Methods: M8L

Pecewse Lnearzaton: Numercal Examle contd. ( ) g x can be wrtten as, g 5 ( x ) w g ( t ) = = w 0 + w + w3 8 + w4 8 + w5 3 4 Thus, the lnear aroxmaton of the above roblem becomes Maxmze f = w 8 w + x subject to w w w + 8w + w + w3 + 7w4 + 64 3 + w + 8w 3 4 + w 4 0 for =,,...,5 + 3w + w 5 5 + x = 5 + s Otmzaton Methods: M8L = 5

Pecewse Lnearzaton: Numercal Examle contd. Ths can be solved usng smlex method n a restrcted bass condton The smlex tableau s shown below 5 Otmzaton Methods: M8L

Pecewse Lnearzaton: Numercal Examle contd. From the table, t s clear that w 5 should be the enterng varable s should be the extng varable But accordng to restrcted bass condton and cannot occur together n bass as they are not adjacent Therefore, consder the next best enterng varable Ths also s not ossble, snce should be exted and and cannot occur together The next best varable, t s clear that should be the extng varable w w5 w 4 s w4 w w w3 6 Otmzaton Methods: M8L

Pecewse Lnearzaton: Numercal Examle contd. The smlex tableau s shown below The enterng varable s w 5. Then the varable to be exted s s and ths s not accetable snce s not an adjacent ont to w w 4 w5 3 Next varable can be admtted by drong. s 7 Otmzaton Methods: M8L

Pecewse Lnearzaton: Numercal Examle contd. The smlex tableau s shown below w w 5 4 w w w3 Now, cannot be admtted snce cannot be droed Smlarly and cannot be entered as cannot be droed 8 Otmzaton Methods: M8L

Pecewse Lnearzaton: Numercal Examle contd. Snce there s no more varable to be entered, the rocess ends Therefore, the best soluton s Now, w3 = 0.3; w4 = 0.7 x = w t = w3 + w4 3 =.7 The otmum value s 5 and x = 0 f =. 3 Ths may be an aroxmate soluton to the orgnal nonlnear roblem However, the soluton can be mroved by takng fner breakng onts 9 Otmzaton Methods: M8L

Thank You 0 Otmzaton Methods: M8L