Single Variable Optimization

Similar documents
Chapter Newton s Method

Summary with Examples for Root finding Methods -Bisection -Newton Raphson -Secant

: Numerical Analysis Topic 2: Solution of Nonlinear Equations Lectures 5-11:

Lecture 2 Solution of Nonlinear Equations ( Root Finding Problems )

OPTIMISATION. Introduction Single Variable Unconstrained Optimisation Multivariable Unconstrained Optimisation Linear Programming

CHAPTER 4d. ROOTS OF EQUATIONS

Chapter 3 Differentiation and Integration

CISE301: Numerical Methods Topic 2: Solution of Nonlinear Equations

36.1 Why is it important to be able to find roots to systems of equations? Up to this point, we have discussed how to find the solution to

Numerical Methods Solution of Nonlinear Equations

Root Finding

Newton s Method for One - Dimensional Optimization - Theory

Review of Taylor Series. Read Section 1.2

Finite Difference Method

Shuai Dong. Isaac Newton. Gottfried Leibniz

Lecture 21: Numerical methods for pricing American type derivatives

ME 501A Seminar in Engineering Analysis Page 1

ORDINARY DIFFERENTIAL EQUATIONS EULER S METHOD

NUMERICAL DIFFERENTIATION

Curve Fitting with the Least Square Method

Work is the change in energy of a system (neglecting heat transfer). To examine what could

Mathematical Economics MEMF e ME. Filomena Garcia. Topic 2 Calculus

Kernel Methods and SVMs Extension

Lecture 8 Optimization

Topic 5: Non-Linear Regression

CHAPTER 7 CONSTRAINED OPTIMIZATION 2: SQP AND GRG

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

= 1.23 m/s 2 [W] Required: t. Solution:!t = = 17 m/s [W]! m/s [W] (two extra digits carried) = 2.1 m/s [W]

Lecture 13 APPROXIMATION OF SECOMD ORDER DERIVATIVES

Polynomial Regression Models

ACTM State Calculus Competition Saturday April 30, 2011

A Tale of Friction Basic Rollercoaster Physics. Fahrenheit Rollercoaster, Hershey, PA max height = 121 ft max speed = 58 mph

PHYS 1443 Section 004 Lecture #12 Thursday, Oct. 2, 2014

Chapter 12. Ordinary Differential Equation Boundary Value (BV) Problems

Chapter 8. Potential Energy and Conservation of Energy

Chapter - 2. Distribution System Power Flow Analysis

Mathematics Intersection of Lines

College of Computer & Information Science Fall 2009 Northeastern University 20 October 2009

Section 8.3 Polar Form of Complex Numbers

Generalized Linear Methods

CHAPTER 5 NUMERICAL EVALUATION OF DYNAMIC RESPONSE

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

Modeling motion with VPython Every program that models the motion of physical objects has two main parts:

Chapter 4: Root Finding

Propagation of error for multivariable function

Solution of Linear System of Equations and Matrix Inversion Gauss Seidel Iteration Method

Which Separator? Spring 1

CS 331 DESIGN AND ANALYSIS OF ALGORITHMS DYNAMIC PROGRAMMING. Dr. Daisy Tang

Maximal Margin Classifier

Numerical Differentiation

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

Singular Value Decomposition: Theory and Applications

Term Project - select journal paper and outline. Completed analysis due end

6.1 The function can be formulated as a fixed-point iteration as

Supplement: Proofs and Technical Details for The Solution Path of the Generalized Lasso

ADVANCED MACHINE LEARNING ADVANCED MACHINE LEARNING

Logistic Regression Maximum Likelihood Estimation

For now, let us focus on a specific model of neurons. These are simplified from reality but can achieve remarkable results.

Lecture 26 Finite Differences and Boundary Value Problems

Integrals and Invariants of Euler-Lagrange Equations

Physics 2A Chapter 3 HW Solutions

Week3, Chapter 4. Position and Displacement. Motion in Two Dimensions. Instantaneous Velocity. Average Velocity

Gravitational Acceleration: A case of constant acceleration (approx. 2 hr.) (6/7/11)

Linear Approximation with Regularization and Moving Least Squares

Lecture Notes on Linear Regression

On the Interval Zoro Symmetric Single-step Procedure for Simultaneous Finding of Polynomial Zeros

2.29 Numerical Fluid Mechanics

PHYS 1441 Section 002 Lecture #15

Section 3.6 Complex Zeros

The Fundamental Theorem of Algebra. Objective To use the Fundamental Theorem of Algebra to solve polynomial equations with complex solutions

Support Vector Machines

Nice plotting of proteins II

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

The equation of motion of a dynamical system is given by a set of differential equations. That is (1)

Lecture 10 Support Vector Machines II

Least squares cubic splines without B-splines S.K. Lucas

Some Comments on Accelerating Convergence of Iterative Sequences Using Direct Inversion of the Iterative Subspace (DIIS)

Computational Biology Lecture 8: Substitution matrices Saad Mneimneh

EEL 6266 Power System Operation and Control. Chapter 3 Economic Dispatch Using Dynamic Programming

PHYS 1101 Practice problem set 12, Chapter 32: 21, 22, 24, 57, 61, 83 Chapter 33: 7, 12, 32, 38, 44, 49, 76

Aerodynamics. Finite Wings Lifting line theory Glauert s method

An Interactive Optimisation Tool for Allocation Problems

IV. Performance Optimization

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

Numerical Methods. ME Mechanical Lab I. Mechanical Engineering ME Lab I

ONE-DIMENSIONAL COLLISIONS

Feature Selection: Part 1

Bezier curves. Michael S. Floater. August 25, These notes provide an introduction to Bezier curves. i=0

Convexity preserving interpolation by splines of arbitrary degree

One-sided finite-difference approximations suitable for use with Richardson extrapolation

Numerical Solution of Ordinary Differential Equations

Line Drawing and Clipping Week 1, Lecture 2

10.34 Numerical Methods Applied to Chemical Engineering Fall Homework #3: Systems of Nonlinear Equations and Optimization

Lecture Torsion Properties for Line Segments and Computational Scheme for Piecewise Straight Section Calculations

Week 9 Chapter 10 Section 1-5

Chapter 5 Function-based Monte Carlo

Physics 2A Chapters 6 - Work & Energy Fall 2017

= = = (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.

Radar Trackers. Study Guide. All chapters, problems, examples and page numbers refer to Applied Optimal Estimation, A. Gelb, Ed.

Difference Equations

Transcription:

8/4/07 Course Instructor Dr. Raymond C. Rump Oce: A 337 Phone: (95) 747 6958 E Mal: rcrump@utep.edu Topc 8b Sngle Varable Optmzaton EE 4386/530 Computatonal Methods n EE Outlne Mathematcal Prelmnares Sngle Varable Optmzaton Parabolc nterpolaton Newton s method Golden secton search D Optmzaton

8/4/07 Mathematcal Prelmnares D Optmzaton 3 Recall Dervatve Tests Frst order dervatves tell us slope and whether we have reached an etremum or not. 0 0 0 0 0 The sgn o the second order dervatves tell us whether our etremum s a mnmum or a mamum. 0 mnmum 0 0 mamum D Optmzaton 4

8/4/07 Sngle Varable Optmzaton: Parabolc Interpolaton D Optmzaton 5 Formulaton o the Method We t () to a polynomal and then use the rst dervatve test to nd the etremum. Step Pck three ponts that span an etremum and and and 3 3 3 Step Ft the ponts to a polynomal a a a 0 a a a a a a a a a 0 0 3 0 3 3 Step 3 se the rst dervatve test to nd etremum a0 a 3 3 3a 0 a a a a e e e 0 3 3 3 a a a Step 4 Ater workng through the algebra, we get a nal epresson or the etremum 3 3 3 e 3 3 3 D Optmzaton 6 a a 3

8/4/07 Vsualzaton o the Method Appromated Etremum True Etremum e 3 D Optmzaton 7 Sngle Varable Optmzaton: Newton s Method D Optmzaton 8 4

8/4/07 Formulaton o the Method Recall that the Newton Raphson method was used to nd the zero o a uncton. Startng wth an ntal guess or the root, the ollowng equaton was terated untl convergence. We can easly convert ths nto an algorthm or ndng an etremum nstead o a zero. Dene an aulary uncton g() that s the rst dervatve o (). d g d The aulary uncton g() wll have a zero at an etremum o (). Ths means we can perorm the Newton Raphson method on g() to nd an etremum o (). g g D Optmzaton 9 Sngle Varable Optmzaton: Golden Secton Search D Optmzaton 0 5

8/4/07 Step Dene Interval We try to pck lower and upper lmts, and, so that () has only one etremum nsde the nterval. D Optmzaton Step Evaluate the Functon at the Bounds D Optmzaton 6

8/4/07 Step 3 Pck Two Intermedate Ponts and d d d d d R D Optmzaton 3 5 R 0.680... Step 4 Evaluate the Functon at Ponts and D Optmzaton 4 7

8/4/07 Step 5a Determne the Poston o the Etremum < Etremum s on rght sde I >, then the etremum s on the let sde o the nterval between and. I <, then the etremum s on the rght sde o the nterval between and. D Optmzaton 5 Step 5b Adjust Ponts R D Optmzaton 6 8

8/4/07 Step 5c Determne the Poston o the Etremum > Etremum s on let sde I >, then the etremum s on the let sde o the nterval between and. I <, then the etremum s on the rght sde o the nterval between and. D Optmzaton 7 Step 5b Adjust Ponts R D Optmzaton 8 9

8/4/07 Step 5c Repeat ntl Convergence Converged : tolerance D Optmzaton 9 Step 6 Calculate Fnal Answer We estmate the nal etremum to be at the mdpont o the last nterval. e D Optmzaton 0 0

8/4/07 Algorthm Summary. Dene startng bounds and.. Evaluate the uncton at the two boundng ponts. 3. Pck two ntermedate ponts, and, usng Golden rato. 5 R 0.68033988749895... R R 4. Evaluate uncton at the two ntermedate ponts. 5. pdate bounds by dentyng poston o the etrema. I >, etrema s on let sde. R I <, etrema s on rght sde. R 6. Repeat Step 5 untl convergence tolerance 7. Calculate nal answer e e e D Optmzaton Anmaton o the Method D Optmzaton

8/4/07 Dervaton o Golden Rato ( o ) From ths pcture, we dene three length parameters. 0 0 Recognzng that we want the ponts on the net teraton to le on top o ponts rom the prevous teraton, we dene two condtons to ensure ths. Condton ensures l + l covers entre span. 0 0 Condton ensures the net teraton has the same proportonal spacng as the current teraton. D Optmzaton 0 3 Dervaton o Golden Rato ( o ) 0 0 Condton ensures l + l covers entre span. Condton ensures the net teraton has the same proportonal spacng as the current teraton. Substtute Condton nto Condton to elmnate l 0. 0 Dene the Golden rato as R = l /l. R R R R 0 We solve or R usng the quadratc ormula. b b 4ac 4 5 R a We pck the postve root to keep R postve. 5 0.68033988749895... D Optmzaton 4 R

8/4/07 The Magc o the Golden Secton Search Iteraton Iteraton + Three o the our ponts lne up rom one teraton to the net. Ths means the uncton only has to be evaluated at one new pont each teraton. The nterval s reduced by 38.% each teraton. D Optmzaton 5 3