GSLM Operations Research II Fall 13/14

Similar documents
The Greedy Method. Outline and Reading. Change Money Problem. Greedy Algorithms. Applications of the Greedy Strategy. The Greedy Method Technique

5 The Primal-Dual Method

Greedy Technique - Definition

Optimization Methods: Integer Programming Integer Linear Programming 1. Module 7 Lecture Notes 1. Integer Linear Programming

Sum of Linear and Fractional Multiobjective Programming Problem under Fuzzy Rules Constraints

LECTURE NOTES Duality Theory, Sensitivity Analysis, and Parametric Programming

Support Vector Machines

Support Vector Machines

Solving two-person zero-sum game by Matlab

Kent State University CS 4/ Design and Analysis of Algorithms. Dept. of Math & Computer Science LECT-16. Dynamic Programming

Classification / Regression Support Vector Machines

11. APPROXIMATION ALGORITHMS

Complex Numbers. Now we also saw that if a and b were both positive then ab = a b. For a second let s forget that restriction and do the following.

Smoothing Spline ANOVA for variable screening

Programming in Fortran 90 : 2017/2018

S1 Note. Basis functions.

NUMERICAL SOLVING OPTIMAL CONTROL PROBLEMS BY THE METHOD OF VARIATIONS

Machine Learning. Support Vector Machines. (contains material adapted from talks by Constantin F. Aliferis & Ioannis Tsamardinos, and Martin Law)

Radial Basis Functions

Problem Set 3 Solutions

Fixing Max-Product: Convergent Message Passing Algorithms for MAP LP-Relaxations

6.854 Advanced Algorithms Petar Maymounkov Problem Set 11 (November 23, 2005) With: Benjamin Rossman, Oren Weimann, and Pouya Kheradpour

An Optimal Algorithm for Prufer Codes *

U.C. Berkeley CS294: Beyond Worst-Case Analysis Handout 5 Luca Trevisan September 7, 2017

Parallelism for Nested Loops with Non-uniform and Flow Dependences

Today Using Fourier-Motzkin elimination for code generation Using Fourier-Motzkin elimination for determining schedule constraints

UNIT 2 : INEQUALITIES AND CONVEX SETS


An Application of the Dulmage-Mendelsohn Decomposition to Sparse Null Space Bases of Full Row Rank Matrices

Meta-heuristics for Multidimensional Knapsack Problems

Polyhedral Compilation Foundations

Optimal Workload-based Weighted Wavelet Synopses

Harmonic Coordinates for Character Articulation PIXAR

Assignment # 2. Farrukh Jabeen Algorithms 510 Assignment #2 Due Date: June 15, 2009.

OPL: a modelling language

Ecient Computation of the Most Probable Motion from Fuzzy. Moshe Ben-Ezra Shmuel Peleg Michael Werman. The Hebrew University of Jerusalem

LP Decoding. Martin J. Wainwright. Electrical Engineering and Computer Science UC Berkeley, CA,

Solitary and Traveling Wave Solutions to a Model. of Long Range Diffusion Involving Flux with. Stability Analysis

Intra-Parametric Analysis of a Fuzzy MOLP

Module Management Tool in Software Development Organizations

12/2/2009. Announcements. Parametric / Non-parametric. Case-Based Reasoning. Nearest-Neighbor on Images. Nearest-Neighbor Classification

Mathematics 256 a course in differential equations for engineering students

Subspace clustering. Clustering. Fundamental to all clustering techniques is the choice of distance measure between data points;

For instance, ; the five basic number-sets are increasingly more n A B & B A A = B (1)

Array transposition in CUDA shared memory

Multicriteria Decision Making

Machine Learning 9. week

Modeling and Solving Nontraditional Optimization Problems Session 2a: Conic Constraints

Machine Learning: Algorithms and Applications

An Application of Network Simplex Method for Minimum Cost Flow Problems

Analysis of Continuous Beams in General

Minimization of the Expected Total Net Loss in a Stationary Multistate Flow Network System

TPL-Aware Displacement-driven Detailed Placement Refinement with Coloring Constraints

The Codesign Challenge

Computer Animation and Visualisation. Lecture 4. Rigging / Skinning

Compiler Design. Spring Register Allocation. Sample Exercises and Solutions. Prof. Pedro C. Diniz

Support Vector Machines. CS534 - Machine Learning

LECTURE : MANIFOLD LEARNING

SENSITIVITY ANALYSIS IN LINEAR PROGRAMMING USING A CALCULATOR

Adaptive Weighted Sum Method for Bi-objective Optimization

Design for Reliability: Case Studies in Manufacturing Process Synthesis

A Binarization Algorithm specialized on Document Images and Photos

Lecture 5: Multilayer Perceptrons

Quality Improvement Algorithm for Tetrahedral Mesh Based on Optimal Delaunay Triangulation

Clustering on antimatroids and convex geometries

A Saturation Binary Neural Network for Crossbar Switching Problem

LOOP ANALYSIS. The second systematic technique to determine all currents and voltages in a circuit

Exercises (Part 4) Introduction to R UCLA/CCPR. John Fox, February 2005

Barycentric Coordinates. From: Mean Value Coordinates for Closed Triangular Meshes by Ju et al.

Announcements. Supervised Learning

CHAPTER 3 SEQUENTIAL MINIMAL OPTIMIZATION TRAINED SUPPORT VECTOR CLASSIFIER FOR CANCER PREDICTION

An Indian Journal FULL PAPER ABSTRACT KEYWORDS. Trade Science Inc.

CS246: Mining Massive Datasets Jure Leskovec, Stanford University

Fast Computation of Shortest Path for Visiting Segments in the Plane

Loop Transformations, Dependences, and Parallelization

Outline. Self-Organizing Maps (SOM) US Hebbian Learning, Cntd. The learning rule is Hebbian like:

Wavefront Reconstructor

A Fuzzy Goal Programming Approach for a Single Machine Scheduling Problem

Introduction to Geometrical Optics - a 2D ray tracing Excel model for spherical mirrors - Part 2

Performance Evaluation of Information Retrieval Systems

All-Pairs Shortest Paths. Approximate All-Pairs shortest paths Approximate distance oracles Spanners and Emulators. Uri Zwick Tel Aviv University

Algorithm To Convert A Decimal To A Fraction

XLVII SIMPÓSIO BRASILEIRO DE PESQUISA OPERACIONAL

CMPS 10 Introduction to Computer Science Lecture Notes

Combinatorial Auctions with Structured Item Graphs

CONCURRENT OPTIMIZATION OF MULTI RESPONCE QUALITY CHARACTERISTICS BASED ON TAGUCHI METHOD. Ümit Terzi*, Kasım Baynal

Cable optimization of a long span cable stayed bridge in La Coruña (Spain)

Inverse Kinematics (part 2) CSE169: Computer Animation Instructor: Steve Rotenberg UCSD, Spring 2016

A Facet Generation Procedure. for solving 0/1 integer programs

SUMMARY... I TABLE OF CONTENTS...II INTRODUCTION...

Lecture 5: Probability Distributions. Random Variables

New Bundle Methods for Solving Lagrangian Relaxation Dual Problems 1

A Fast Content-Based Multimedia Retrieval Technique Using Compressed Data

Biostatistics 615/815

Scheduling with Integer Time Budgeting for Low-Power Optimization

(1) The control processes are too complex to analyze by conventional quantitative techniques.

$OJRULWKPV. (Feodor F. Dragan) Department of Computer Science Kent State University

ELEC 377 Operating Systems. Week 6 Class 3

Explicit Formulas and Efficient Algorithm for Moment Computation of Coupled RC Trees with Lumped and Distributed Elements

Overlapping Clustering with Sparseness Constraints

Transcription:

GSLM 58 Operatons Research II Fall /4 6. Separable Programmng Consder a general NLP mn f(x) s.t. g j (x) b j j =. m. Defnton 6.. The NLP s a separable program f ts objectve functon and all constrants are conssted of separable functons.e. f(x) = n f ( x ) and n g ( x ) b j = m; j j and all x are non-negatve varables bounded above.e. x for some = n. The technque separable programmng bascally replaces all separable functons n objectves and constrants by pecewse lnear functons. Defnton 6.. A convex program s an NLP that mnmzes a convex functon or maxmzes a concave functon over a convex set. Fact 6.. Any (contnuous) convex functon can be approxmated to any degree of accuracy by a pecewse lnear convex functon. From Fact 6. when f j and g j are convex functons they can be approxmated to any degree of accuracy by pecewse lnear functons. Eventually such an NLP of pecewse convex functons can be represented by a lnear program (LP). Thus effectvely a separable convex program can be approxmated by a sequence of LPs to any degree of accuracy. Note also that when f j and g j are convex an local mnmum s n fact a global mnmum. Example 6.. NLP s: mn x x x s.t. x + x 5 x + x 9 x x. The above s a separable program wth f (x ) =.5 6 9 4 O A 4 B C

GSLM 58 Operatons Research II Fall /4 x x and f (x ) = -x. f s lnear. The problem s easy to solve f f.e. x s approxmated by a lnear functon. From the constrants x 4.5. Let us approxmate the non-lnear functon y = x by a pecewse lnear functon. For smplcty we take a functon of three lnear peces wth break ponts at 4 and 4.5. The same procedure can be appled to any number of break ponts at any values. ponts O A B C x 4 4.5 y 4 6.5 There are two ways the - and the -forms to represent the functon n pecewse lnear form. Defnton 6.. (-form) For any pont wthn a lnear segment ts functonal value s the convex combnaton of the values of the two break ponts of the lnear segment. Let be the weght of break pont = O A B and C. x O A 4B 4.5C y O 4 A 6B.5C O A B C at most two adjacent take non - zero values. () () () (4) Now reformulate NLP nto NLP mn y x x s.t. x + x 5 x + x 9 () () () and (4) x x = O A B C.

GSLM 58 Operatons Research II Fall /4 Constrants (4) can be omtted f NLP s a convex program. Check that whenever (4) s volated by a set of the value of the correspondng y s above the three-pece lnear functon and hence the set of cannot be a mnmum pont. In general f the NLP s a non-convex program constrants (4) are needed though they are mplemented mplctly through the separable programmng extenson of the smplex method. The pvotng step of the separable programmng extenson smply ensures that at any tme at most two adjacent are n the bass. It can be shown that such a restrcted bass rule wll lead to the optmum soluton. The dea of separable program s applcable to non-convex programs. Of course n those cases the optmum soluton can be a local rather than global optmum. Example 6.. (Non-Convex problem) mn f(x) s.t. x. f(x) 6 f(x) s a pecewse lnear functon (whch s not lnear by tself). It s obvous that mnmum s x * = wth f(x * ) =. Let us fnd ths by the dea of separable programmng. x Let x = + A + B and y = + A + 6 B ; and add n the constrant + A + B =. We take the specal attenton of (4) that at most two s can be postve at any tme. The problem becomes mn + A + 6 B s.t. + A + B + A + B + A + B = A B and at most two adjacent s can be postve at the same tme. After addng n slack varable s surplus varable u and two artfcal varables a and a the problem becomes mn + A + 6 B + Ma + Ma

GSLM 58 Operatons Research II Fall /4 s.t. + A + B + s = + A + B u + a = + A + B + a = A B s a a and at most two adjacent s can be postve at the same tme. A B s u a a RHS 6 M M s a - a A B s u a a RHS M M 64M M M s a - a A B s u a a RHS 5M 8 M 64M M s B a Supposedly s the most negatve. However snce B s n the bass only A among the s s qualfed to be n the bass. A B s u a a RHS 8 M 8M M M M M s A a 4

GSLM 58 Operatons Research II Fall /4 A B s u a a RHS 8+M +M M -+M s - - a - - - A B s u a a RHS 9 5 M s - u - - - Defnton 6.4. (-form) Instead of takng weghted value of break ponts we can add up the contrbuton from each lnear segment. Let be the proporton of the th segment taken =. there exsts j x y 4 such that.5 4.5 for j and for j. (5) (6) (7) (8) As for the -form the -form may also gve a local optmum f the orgnal program s non-convex. When the orgnal program s convex constrant (8) s not necessary. Remark 6.. To get more accurate result the pecewse lnear approxmaton of f can be refned wth more lnear segments. There are studes on segment refnement to get the best trade off between accuracy and computatonal effort. Remark 6.. It s possble to approxmate constrants by smlar procedure. Remark 6.. A product term x x can be transformed to separable form. Let s = (x +x )/ s and s = (x -x )/. Then x x = s. 5