an introduction to optimization theory. outline introduction unconstrained optimization problem...

30
An Introduction to Optimization Theory

Post on 22-Dec-2015

274 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: An Introduction to Optimization Theory. Outline Introduction Unconstrained optimization problem Constrained optimization problem

An Introduction to Optimization Theory

Page 2: An Introduction to Optimization Theory. Outline Introduction Unconstrained optimization problem Constrained optimization problem

Outline

Introduction

Unconstrained optimization problem

Constrained optimization problem

Page 3: An Introduction to Optimization Theory. Outline Introduction Unconstrained optimization problem Constrained optimization problem

Introduction

Mathematically speaking, optimization is the minimization of a objective function subject to constraints on its variables. Mathematically, we have

set) (Feasible choices possible all ofset theis

function objective is )( where

subject to

)(minarg

xf

x

xfx

Page 4: An Introduction to Optimization Theory. Outline Introduction Unconstrained optimization problem Constrained optimization problem

Introduction

Page 5: An Introduction to Optimization Theory. Outline Introduction Unconstrained optimization problem Constrained optimization problem

Introduction-Linear regression

Page 6: An Introduction to Optimization Theory. Outline Introduction Unconstrained optimization problem Constrained optimization problem

Introduction-Battery charger

Page 7: An Introduction to Optimization Theory. Outline Introduction Unconstrained optimization problem Constrained optimization problem

Unconstrained optimization problem

Definition for unconstrained optimization problem:

Page 8: An Introduction to Optimization Theory. Outline Introduction Unconstrained optimization problem Constrained optimization problem

Unconstrained optimization problem

Page 9: An Introduction to Optimization Theory. Outline Introduction Unconstrained optimization problem Constrained optimization problem

Gradient descent algorithm

Page 10: An Introduction to Optimization Theory. Outline Introduction Unconstrained optimization problem Constrained optimization problem

Gradient descent algorithm

Gradient descent algorithm may be trapped into the local extreme instead of the global extreme

Page 11: An Introduction to Optimization Theory. Outline Introduction Unconstrained optimization problem Constrained optimization problem

Gradient descent algorithm

Methodology for choosing suitable step size αk ---- Steepest descent algorithm

Page 12: An Introduction to Optimization Theory. Outline Introduction Unconstrained optimization problem Constrained optimization problem

Gradient descent algorithm

Page 13: An Introduction to Optimization Theory. Outline Introduction Unconstrained optimization problem Constrained optimization problem

Gradient descent algorithm

Steepest descent algorithm with quadratic cost function:

Page 14: An Introduction to Optimization Theory. Outline Introduction Unconstrained optimization problem Constrained optimization problem

Gradient descent algorithm

bQxxfg kkk )()()( )(Update equation:

Page 15: An Introduction to Optimization Theory. Outline Introduction Unconstrained optimization problem Constrained optimization problem

Newton method

Summary for Newton method

Page 16: An Introduction to Optimization Theory. Outline Introduction Unconstrained optimization problem Constrained optimization problem

Newton method

Page 17: An Introduction to Optimization Theory. Outline Introduction Unconstrained optimization problem Constrained optimization problem

Newton method

Procedure for Newton method

Page 18: An Introduction to Optimization Theory. Outline Introduction Unconstrained optimization problem Constrained optimization problem

Quasi-Newton method

Page 19: An Introduction to Optimization Theory. Outline Introduction Unconstrained optimization problem Constrained optimization problem

Quasi-Newton method

What properties of F(x(k))-1 should it mimic ?

1. Hk should be a symmetric matrix

2. Hk should with secant property

)1()(

)1()()( )(')(')(''

kk

kkk

xx

xfxfxf

Page 20: An Introduction to Optimization Theory. Outline Introduction Unconstrained optimization problem Constrained optimization problem

Quasi-Newton method

Typical approaches for Quasi-Newton method

1. Rank-one formula

2. DFP algorithm

3. BFGS algorithm (L-BFGS , L indicates limited-memory)

Page 21: An Introduction to Optimization Theory. Outline Introduction Unconstrained optimization problem Constrained optimization problem

Constrained optimization problem

Definition for constrained optimization problem

Page 22: An Introduction to Optimization Theory. Outline Introduction Unconstrained optimization problem Constrained optimization problem

Problems with equality constraints ---- Lagrange multiplier

Page 23: An Introduction to Optimization Theory. Outline Introduction Unconstrained optimization problem Constrained optimization problem

Problems with equality constraints ---- Lagrange multiplier

Page 24: An Introduction to Optimization Theory. Outline Introduction Unconstrained optimization problem Constrained optimization problem

Problems with equality constraints ---- Lagrange multiplier

Page 25: An Introduction to Optimization Theory. Outline Introduction Unconstrained optimization problem Constrained optimization problem

Problems with equality constraints ---- Lagrange multiplier

Suppose x* is a local minimizer

Page 26: An Introduction to Optimization Theory. Outline Introduction Unconstrained optimization problem Constrained optimization problem
Page 27: An Introduction to Optimization Theory. Outline Introduction Unconstrained optimization problem Constrained optimization problem

Karush-Kuhn-Tucker condition (KKT)

From now on, we will consider the following problem

Page 28: An Introduction to Optimization Theory. Outline Introduction Unconstrained optimization problem Constrained optimization problem

Karush-Kuhn-Tucker condition (KKT)

Note that:

Page 29: An Introduction to Optimization Theory. Outline Introduction Unconstrained optimization problem Constrained optimization problem

Image statistics & Image enhancement

Illustration for gradient descent with projection

operator Projection:

][][ :equation Projection

:equationdescent Gradient )()1()1(

)()1(

kkkkk

p

kkkk

dxxx

dxx

Constrained set Ω

Initial solution

Projection

Page 30: An Introduction to Optimization Theory. Outline Introduction Unconstrained optimization problem Constrained optimization problem

Useful Matlab introductions for optimization

Useful instructions included in Matlab for optimization

1. fminunc: Solver for unconstrained optimization problems

2. fmincon: Solver for constrained optimization problems

3. linprog: Solver for linear programming problems

4. quadprog: Solver for quadratic programming problems