i.e., into a monomial, using the Arithmetic-Geometric Mean Inequality, the result will be a posynomial approximation!

Similar documents
Unconstrained Geometric Programming

ELE539A: Optimization of Communication Systems Lecture 16: Pareto Optimization and Nonconvex Optimization

Numerical Optimization

Introduction to Mathematical Programming IE406. Lecture 10. Dr. Ted Ralphs

Benders Decomposition

12. Interior-point methods

Following The Central Trajectory Using The Monomial Method Rather Than Newton's Method

Primal-Dual Interior-Point Methods for Linear Programming based on Newton s Method

Algorithms and Theory of Computation. Lecture 13: Linear Programming (2)

Interior Point Algorithms for Constrained Convex Optimization

2001 Dennis L. Bricker Dept. of Industrial Engineering The University of Iowa. Reducing dimensionality of DP page 1

2018 년전기입시기출문제 2017/8/21~22

(includes both Phases I & II)

Linear Programming. Operations Research. Anthony Papavasiliou 1 / 21

Sensitivity Analysis

Convex Optimization. Newton s method. ENSAE: Optimisation 1/44

Lecture 5. x 1,x 2,x 3 0 (1)

Barrier Method. Javier Peña Convex Optimization /36-725

Lecture 11: Post-Optimal Analysis. September 23, 2009

Dual Basic Solutions. Observation 5.7. Consider LP in standard form with A 2 R m n,rank(a) =m, and dual LP:

12. Interior-point methods

A Review of Linear Programming

(includes both Phases I & II)

Example Problem. Linear Program (standard form) CSCI5654 (Linear Programming, Fall 2013) Lecture-7. Duality

Semidefinite Programming

3.10 Lagrangian relaxation

1 Number Systems and Errors 1

Benders Decomposition Methods for Structured Optimization, including Stochastic Optimization

Summary of the simplex method

maxz = 3x 1 +4x 2 2x 1 +x 2 6 2x 1 +3x 2 9 x 1,x 2

Facial Reduction and Geometry on Conic Programming

The Concept of Neutrosophic Less Than or Equal To: A New Insight in Unconstrained Geometric Programming

This research was partially supported by the Faculty Research and Development Fund of the University of North Carolina at Wilmington

Lagrange duality. The Lagrangian. We consider an optimization program of the form

Review Solutions, Exam 2, Operations Research

CS711008Z Algorithm Design and Analysis

Unit 3 Vocabulary. An algebraic expression that can contains. variables, numbers and operators (like +, An equation is a math sentence stating

Linear Programming: Chapter 5 Duality

Lecture 9 Sequential unconstrained minimization

Optimization for Communications and Networks. Poompat Saengudomlert. Session 4 Duality and Lagrange Multipliers

Benders' Method Paul A Jensen

LINEAR PROGRAMMING 2. In many business and policy making situations the following type of problem is encountered:

The Simplex Algorithm

Today: Linear Programming (con t.)

Lecture 13: Constrained optimization

The dual simplex method with bounds

4. Algebra and Duality

Introduction to Applied Linear Algebra with MATLAB

Adding and Subtracting Polynomials Add and Subtract Polynomials by doing the following: Combine like terms

20A. Build. Build and add. Build a rectangle and find the area (product). l e s s o n p r a c t i c e 1. X X X 2 + 6X X

Discrete (and Continuous) Optimization WI4 131

Research Article A New Global Optimization Algorithm for Solving Generalized Geometric Programming

Duality revisited. Javier Peña Convex Optimization /36-725

Part 1. The Review of Linear Programming

LP Duality: outline. Duality theory for Linear Programming. alternatives. optimization I Idea: polyhedra

Example: feasibility. Interpretation as formal proof. Example: linear inequalities and Farkas lemma

Motivation. Lecture 2 Topics from Optimization and Duality. network utility maximization (NUM) problem:

- Well-characterized problems, min-max relations, approximate certificates. - LP problems in the standard form, primal and dual linear programs

Lecture 15: October 15

4. Convex optimization problems

ORIE 6300 Mathematical Programming I August 25, Lecture 2

Convex optimization problems. Optimization problem in standard form

Lecture #21. c T x Ax b. maximize subject to

Lecture 9 Tuesday, 4/20/10. Linear Programming

The most factored form is usually accomplished by common factoring the expression. But, any type of factoring may come into play.

CSCI 1951-G Optimization Methods in Finance Part 09: Interior Point Methods

Approximate Farkas Lemmas in Convex Optimization

Pacing Guide Algebra 1

Lecture: Convex Optimization Problems

A Strongly Polynomial Simplex Method for Totally Unimodular LP

Duality (Continued) min f ( x), X R R. Recall, the general primal problem is. The Lagrangian is a function. defined by

Midterm Review. Yinyu Ye Department of Management Science and Engineering Stanford University Stanford, CA 94305, U.S.A.

(P ) Minimize 4x 1 + 6x 2 + 5x 3 s.t. 2x 1 3x 3 3 3x 2 2x 3 6

Convex Optimization and SVM

EE364a Homework 8 solutions

Section Notes 9. IP: Cutting Planes. Applied Math 121. Week of April 12, 2010

Farkas Lemma, Dual Simplex and Sensitivity Analysis

4-1 The Algebraic-Geometric Dictionary P. Parrilo and S. Lall, ECC

Interior-Point Methods for Linear Optimization

I.3. LMI DUALITY. Didier HENRION EECI Graduate School on Control Supélec - Spring 2010

SOLVING INTEGER LINEAR PROGRAMS. 1. Solving the LP relaxation. 2. How to deal with fractional solutions?

Lecture 10. ( x domf. Remark 1 The domain of the conjugate function is given by

2.098/6.255/ Optimization Methods Practice True/False Questions

Primal-Dual Interior-Point Methods. Javier Peña Convex Optimization /36-725

Integer Programming ISE 418. Lecture 8. Dr. Ted Ralphs

POLYNOMIAL OPTIMIZATION WITH SUMS-OF-SQUARES INTERPOLANTS

Discrete Optimization

MVE165/MMG631 Linear and integer optimization with applications Lecture 8 Discrete optimization: theory and algorithms

MVE165/MMG631 Linear and integer optimization with applications Lecture 5 Linear programming duality and sensitivity analysis

Section Notes 9. Midterm 2 Review. Applied Math / Engineering Sciences 121. Week of December 3, 2018

Agenda. 1 Duality for LP. 2 Theorem of alternatives. 3 Conic Duality. 4 Dual cones. 5 Geometric view of cone programs. 6 Conic duality theorem

Sum-Power Iterative Watefilling Algorithm

1. Introduction. A signomial is a weighted sum of exponentials composed with linear functionals of a variable x R n :

CSCI 1951-G Optimization Methods in Finance Part 01: Linear Programming

Feasibility Pump Heuristics for Column Generation Approaches

Linear Programming: Simplex

Section 2.1: Reduce Rational Expressions

Machine Learning. Support Vector Machines. Manfred Huber

ACM 113 Introduction to Optimization - Final Exam. June 6, 2001

ICS-E4030 Kernel Methods in Machine Learning

Transcription:

Dennis L. Bricker Dept of Mechanical & Industrial Engineering The University of Iowa i.e., 1 1 1 Minimize X X X subject to XX 4 X 1 0.5X 1 Minimize X X X X 1X X s.t. 4 1 1 1 1 4X X 1 1 1 1 0.5X X X 1 1 1 1 4/11/006 page 1 of 51 4/11/006 page of 51 Introduce a new variable X so that the signomial now appears in a constraint: Minimize X subject to X X X 4 X 1X X X 1 1 1 1 4X X 1 1 1 1 0.5X X X 1 1 1 1 Next, rewrite the signomial constraint: X X 1 X X X X X 4 1 1 1 1 X X 1 X X X X X 4 1 1 1 1 If we condense the denominator X X X X 1 1 1 into a monomial, using the Arithmetic-Geometric Mean Inequality, the result will be a posynomial approximation! 4/11/006 page of 51 4/11/006 page 4 of 51

u n n i ui i1 i1 i n for all satisfying i1 with equality if & only if 1 1, 0 i u1 u un i i n Condensing the numerator: for such that 1 X X1 X X1 1 1 1 X X X X 1 1, i 0, i 1,, X X X X 1 1 X X X 1 1 1 1 coefficient C 4/11/006 page 5 of 51 4/11/006 page 6 of 51 We choose so that, for a given X, X 1 X X1 X X1 X1 X X X1 X X1 so that 1 1 X 1 X X1 X X1 Minimize X subject to 4 X X1 1 X1 X X X X 1 1 1 1 1 1 1 1 1 4X X 1 0.5X X X 1 and the approximation is exact at X. 4/11/006 page 7 of 51 4/11/006 page 8 of 51

We will choose X,,1 as the initial point. 1 X 1 1 0.047619 1 16 X1 X 16 0.761905 1 4 X1 4 0.190476 1 So we get the posynomial approximation X X 1 X.77409 4 1 1 1.714 0.7619 0.0476 X1 X X which is the posynomial constraint: 1 1 1 1 X X X X X X 1.714 0.7619 0.0476 1 1.77409 0.64965X X X 1.714 1.81 0.0476 1 0.64965X X X.8571 0.7619 0.0476 1 0.64965X X X 1.714 0.7619 0.0476 1 0.64965X X X 1 0.8571 0.7619 0.0476 1 4/11/006 page 9 of 51 4/11/006 page 10 of 51 Posynomial GP approximation of the signomial GP: Minimize X subject to 0.64965X X X 1.714 1.81 0.0476 1 0.64965X X X.8571 0.7619 0.0476 1 0.64965X X X 1.714 0.7619 0.0476 1 0.64965X X X 1 0.8571 0.7619 0.0476 1 4X X 1 1 1 1 0.5X X X 1 1 1 1 4/11/006 page 11 of 51 4/11/006 page 1 of 51

Number of variables: Number of polynomials: 4 Total number of terms: 10 Degrees of difficulty: 6 Terms per polynomial: 1 6 1 Rosenbrock et al. t p Ct exponents 1 1 1 0 0 1 ------------ --------- 1 0 1 1 1 4 1 4 0 1 5 1 0 0 1 6 1 0 1 7 1 0 1 ------------ --------- 8 4 1 1 0 ------------ --------- 9 4 0.5 1 0 10 4 0 1 0 ------------ --------- t = term number, p = polynomial Ct = coefficient 4/11/006 page 1 of 51 4/11/006 page 14 of 51 Bounds on variables # var LB UB 1 X[1] 0.00001 100000 X[] 0.00001 100000 X[] 0.00001 100000 Current Parameters Tolerances for duality gap: 0.0001 Tolerances for constraints (maximum allowable infeasibility) = 0.0001 Tolerance "epsilon" for stopping criterion: epsilon > sum d_rho, where d_rho is the vector of changes in the weights for the terms of the polynomials: epsilon = 0.0005 Maximum # Posynomial subproblems to be solved = 5 Maximum # LPs to be solved per posynomial subproblem = 5 4/11/006 page 15 of 51 4/11/006 page 16 of 51

User-specified Grid Point 1 X[1] X[] The objective function value for this point is 5 The values of the constraint polynomials are: k P(k) 1 1.5 User-specified Grid Point 1 X[1] X[] X[] 1 The objective function value for this point is 1 The values of the constraint polynomials are: k P(k) 5 1 4 1.5 4/11/006 page 17 of 51 4/11/006 page 18 of 51 <>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<> Major Iteration # 1 1 X[1] X[] X[] 1 1 1 1.0 4.0 16.0 4 16.0 5 1.0 6 4.0 7 4.0 8 1.0 9 4 0.5 10 4 1.0 constraint Value Infeasibility 5.0 4.0 1.0 0.0 4 1.5 0.5 poly term value 0.761905 5 0.190476 Objective function = 1 4/11/006 page 19 of 51 4/11/006 page 0 of 51

Condensation of Signomial GP Number of variables: Number of posynomials: 10 Total number of terms: 14 Degrees of difficulty: 11 Terms per posynomial: 1 4 1 1 1 1 1 1 1 (includes bounds on variables to ensure dual feasibility) 0.64965 1.7149 1.81 0.047619 0.64965.8571 0.761905 0.047619 4 0.64965 1.7149 0.761905 0.047619 5 0.64965 0.85714 0.761905 0.047619 6 4 1 1 0 7 4 0.5 1 0 8 4 0 1 0 1 8 0.00001 1 0 0 1 9 0.00001 0 1 0 4/11/006 page 1 of 51 4/11/006 page of 51 1 X[1] 1.597846 X[].66614 X[] 0.6564 Primal: 0.6564 Dual: 0.6564 <>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<> Major Iteration # 1 X[1] 1.597846 X[].66614 X[] 0.6564 1 1 0.6564 19.0171 6.8048 4 17.8718 5.74911 6 8.799 7 6.985 8 0.949464 9 4 0.408 10 4 0.758549 Objective function = 0.6564 4/11/006 page of 51 4/11/006 page 4 of 51

0.996581 0.000000000 4.0059 0.949464 0.000000000 0.0000 4 1.00061 0.0006115 0.0000 small infeasibility! poly term value change 0.790811 0.0890610 5 0.18771 0.007645 Condensation of Signomial GP 0.8551 1.769 1.0919 0.014775 0.8551.067 0.790811 0.014775 4 0.8551 1.769 0.790811 0.014775 5 0.8551 0.0667 0.790811 0.014775 6 4 1 1 0 7 4 0.5 1 0 8 4 0 1 0 1 8 0.00001 1 0 0 1 9 0.00001 0 1 0 4/11/006 page 5 of 51 4/11/006 page 6 of 51 1 X[1] 1.561004 X[].60590 X[] 0.8918 Primal: 0.8918 Dual: 0.8918 <>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<> Major Iteration # 1 X[1] 1.561004 X[].60590 X[] 0.8918 1 1 0.8918.48061 4.91675 4 0.5668 5.4580 6 10.796007 7 8.4606 8 0.9816 9 4 0.768 10 4 0.767480 Objective function = 0.8918 4/11/006 page 7 of 51 4/11/006 page 8 of 51

1.1876 0.18761.4648 large infeasibility! 0.9816 0.00000000 0.0000 4 1.00148 0.001487 46.5604 poly term value change 0.78871 0.005994 5 0.19780 0.00606819 Condensation of Signomial GP 0.84107 1.770 1.117 0.017949 0.84107.968 0.78871 0.017949 4 0.84107 1.770 0.78871 0.017949 5 0.84107 0.9678 0.78871 0.017949 6 4 1 1 0 7 4 0.5 1 0 8 4 0 1 0 1 8 0.00001 1 0 0 1 9 0.00001 0 1 0 4/11/006 page 9 of 51 4/11/006 page 0 of 51 1 X[1] 1.54998 X[].600550 X[] 0.470 Primal: 0.470 Dual: 0.470 <>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<> Major Iteration # 4 1 X[1] 1.54998 X[].600550 X[] 0.470 1 1 0.470 19.47818 5.986660 4 16.61670 5.880168 6 8.9816 7 6.919047 8 0.9986 9 4 0.094 10 4 0.769068 Objective function = 0.470 4/11/006 page 1 of 51 4/11/006 page of 51

0.98446 0.0000000000 11.9815 0.9986 0.0000000000 0.0000 4 1.000010 0.0000100608 47.088 small infeasibility! poly term value change 0.78770 0.0045009 5 0.194451 0.000670688 Notice that the solution seems to alternate between one with small infeasibility and objective approximately 0.47, and one with large infeasibility and objective approximately 0.84 Condensation of Signomial GP 0.8061 1.76199 1.16 0.017795 0.8061.801 0.7877 0.017795 4 0.8061 1.76199 0.7877 0.017795 5 0.8061 0.8009 0.7877 0.017795 6 4 1 1 0 7 4 0.5 1 0 8 4 0 1 0 1 8 0.00001 1 0 0 1 9 0.00001 0 1 0 4/11/006 page of 51 4/11/006 page 4 of 51 1 X[1] 1.548186 X[].595916 X[] 0.84417 Primal: 0.84417 Dual: 0.84417 <>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<> Major Iteration # 5 1 X[1] 1.548186 X[].595916 X[] 0.84417 1 1 0.84417.698 4.759 4 0.19998 5.515961 6 10.88671 7 8.47 8 0.9958 9 4 0.08 10 4 0.770441 Objective function = 0.84417 4/11/006 page 5 of 51 4/11/006 page 6 of 51

1.195860 0.19585984 4.05055 large infeasibility! 0.9958 0.00000000 0.00000 4 1.0017 0.001779 45.9148 poly term value change 0.78664 0.005980 5 0.195664 0.001198 Condensation of Signomial GP 0.84 1.7689 1.164 0.017977 0.84.161 0.78664 0.017977 4 0.84 1.7689 0.78664 0.017977 5 0.84 0.1609 0.78664 0.017977 6 4 1 1 0 7 4 0.5 1 0 8 4 0 1 0 1 8 0.00001 1 0 0 1 9 0.00001 0 1 0 4/11/006 page 7 of 51 4/11/006 page 8 of 51 1 X[1] 1.549449 X[].600175 X[] 0.47004 Primal: 0.47004 Dual: 0.47004 <>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<> Major Iteration # 6 1 X[1] 1.549449 X[].600175 X[] 0.47004 1 1 0.47004 19.4868 5.9790 4 16.610155 5.881808 6 8.9046 7 6.918618 8 0.9984 9 4 0.080 10 4 0.769179 Objective function = 0.47004 4/11/006 page 9 of 51 4/11/006 page 40 of 51

0.98456 0.0000000000 5.9064 0.9984 0.0000000000 0.00000 4 1.000009 0.000008895 54.008 small infeasibility! poly term value change 0.78696 0.0066745 5 0.1945 0.00114178 Condensation of Signomial GP 0.8058 1.76191 1.16 0.017819 0.8058.809 0.78696 0.017819 4 0.8058 1.76191 0.78696 0.017819 5 0.8058 0.8086 0.78696 0.017819 6 4 1 1 0 7 4 0.5 1 0 8 4 0 1 0 1 8 0.00001 1 0 0 1 9 0.00001 0 1 0 4/11/006 page 41 of 51 4/11/006 page 4 of 51 1 X[1] 1.54807 X[].5958 X[] 0.849 Primal: 0.849 Dual: 0.849 <>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<> Major Iteration # 7 1 X[1] 1.54807 X[].5958 X[] 0.849 1 1 0.8490.6989 4.74974 4 0.195068 5.51601 6 10.886941 7 8.46858 8 0.99594 9 4 0.0805 10 4 0.770468 Objective function = 0.849 4/11/006 page 4 of 51 4/11/006 page 44 of 51

1.195904 0.1959058 4.0566 large infeasibility! 0.99594 0.00000000 0.00000 4 1.0017 0.00170 45.90966 poly term value change 0.78645 0.0064891 5 0.195681 0.00115910 Condensation of Signomial GP 0.8 1.7687 1.165 0.017979 0.8.16 0.78645 0.017979 4 0.8 1.7687 0.78645 0.017979 5 0.8 0.169 0.78645 0.017979 6 4 1 1 0 7 4 0.5 1 0 8 4 0 1 0 1 8 0.00001 1 0 0 1 9 0.00001 0 1 0 4/11/006 page 45 of 51 4/11/006 page 46 of 51 1 X[1] 1.54946 X[].600165 X[] 0.47001 Primal: 0.47001 Dual: 0.47001 <>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<> Major Iteration # 8 1 X[1] 1.54946 X[].600165 X[] 0.47001 1 1 0.47001 19.48698 5.978894 4 16.60979 5.88188 6 8.90448 7 6.918579 8 0.99854 9 4 0.087 10 4 0.76918 Objective function = 0.47001 4/11/006 page 47 of 51 4/11/006 page 48 of 51

0.984567 0.00000000000 5.87 0.99854 0.00000000000 0.00000 4 1.000009 0.0000088645 54.06786 small infeasibility! <>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<> 4/11/006 page 49 of 51 4/11/006 page 50 of 51 Dropping two initial points from the plot: zig-zagging behavior is apparent! 4/11/006 page 51 of 51