hard optimization problems: practical approach

51
Hard Optimization Problems: Practical Approach DORIT RON Tel. 08 934 2141 Ziskind room #303 [email protected]

Upload: tyra

Post on 13-Jan-2016

27 views

Category:

Documents


0 download

DESCRIPTION

Hard Optimization Problems: Practical Approach. DORIT RON Tel. 08 934 2141 Ziskind room #303 [email protected]. Course outline. 1 st lecture: Introduction and motivation. INTRODUCTION. What is an optimization problem?. An optimization problem consist of:. Variables : - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Hard Optimization Problems: Practical Approach

Hard Optimization Problems: Practical

Approach

DORIT RON

Tel. 08 934 2141Ziskind room #303 [email protected]

Page 2: Hard Optimization Problems: Practical Approach

Course outline

1st lecture: Introduction and motivation

Page 3: Hard Optimization Problems: Practical Approach

INTRODUCTION

What is an optimization problem?

Page 4: Hard Optimization Problems: Practical Approach

An optimization problem consist of:

Variables:

Energy functional to be minimized/maximized:

min / max

),...,,( 21 nxxxx

)(xE

Page 5: Hard Optimization Problems: Practical Approach

Unconstrained minimizationFind the global minimum

Page 6: Hard Optimization Problems: Practical Approach

An optimization problem consist of:

Variables:

Energy functional to minimized/maximized:

min / max

Possibly subject to: Equality constraints:

Inequality constraints:

),...,,( 21 nxxxx

)(xE

,0)( xfi

,0)( xfi

mi ,...,1

pi ,...,1

Page 7: Hard Optimization Problems: Practical Approach

Constrained minimizationsubject to 0,0 yx

Page 8: Hard Optimization Problems: Practical Approach

INTRODUCTION

What is an optimization problem? Examples

Page 9: Hard Optimization Problems: Practical Approach

Example 1: 2D Ising spins Discrete (combinatorial) optimization

min -i,j> si sj

si = { +1 , -1}

+ + + + - - + -

+ + + + + - - -

- + + + - - - +

- - + + + + - -

- + + - - - - +

+ + + + + - + +

- - - + + + + +

- - - - - + + -

Page 10: Hard Optimization Problems: Practical Approach

3D Ising model Each spin represents a tiny magnet The spins tend to align below a certain Tc

Ferromagnet – Iron at room temperature

magnet ------ Tc------ non-magnet

----|------------------------|---------------------------|--> T

room temp ferromagnetism melting 770oC 1538oC

At T=0 the system settles at its ground states

Page 11: Hard Optimization Problems: Practical Approach

Example 2: 1D graph ordering Given a graph G=(V,), find a

permutation of the vertices that minimizes

E()= i j wi j | (i) -(j) |p

where i , j are in V and wi j is the edge weight between i and j (wi j =0 if ij is not in )

p=1 : Linear arrangement p=2 : Quadratic energy p= : The Bandwidth

Page 12: Hard Optimization Problems: Practical Approach

ij

ijw

Minimum Linear Arrangement Problem

Page 13: Hard Optimization Problems: Practical Approach

Minimum Linear Arrangement Problem

ij

1 2 3 4 5

ijw

Page 14: Hard Optimization Problems: Practical Approach

Minimum Linear Arrangement Problem

ij

1 2 3 4 5

4)(

2)(

j

i

ijw

Page 15: Hard Optimization Problems: Practical Approach

Minimum Linear Arrangement Problem

ij

ijw

1 2 3 4 5

ijw4)(

2)(

j

i

Page 16: Hard Optimization Problems: Practical Approach

Minimum Linear Arrangement Problem

ij

ijw

1 2 3 4 5

ijw4)(

2)(

j

i

,

minimize over all : ( ) = ( ) ( )iji j

E w i j

Page 17: Hard Optimization Problems: Practical Approach

Minimum Linear Arrangement Problem

ij

ijw

1 2 3 4 5

ijw4)(

2)(

j

i

,

minimize over all : ( ) = ( ) ( )iji j

E w i j

9)( E

Page 18: Hard Optimization Problems: Practical Approach

General Linear Arrangement Problem

Variable nodes sizes

E(x)= i j wi j | xi -xj |p

xi = vi /2 + k:k)<i) vk

i j

xi xj

Page 19: Hard Optimization Problems: Practical Approach

Other graph ordering problems

Minimize various functionals:

envelope size, cutwidth, profile of graph, etc. Traveling salesman problem – TSP

Page 20: Hard Optimization Problems: Practical Approach

The Traveling Salesman Problem

Page 21: Hard Optimization Problems: Practical Approach

The Traveling Salesman Problem

Page 22: Hard Optimization Problems: Practical Approach

Other graph ordering problems

Minimize various functionals:

envelope size, cutwidth, profile of graph, etc. Traveling salesman problem – TSP Graph bisectioning Graph partitioning Graph coloring Graph drawing

Page 23: Hard Optimization Problems: Practical Approach

Drawing Graphs

Page 24: Hard Optimization Problems: Practical Approach

Drawing Graphs

Page 25: Hard Optimization Problems: Practical Approach

Drawing Graphs

Page 26: Hard Optimization Problems: Practical Approach

Example 3: 2D circuit placement Bottleneck in the microchip industry Given a hypergraph, find the discrete

placement of each module (gate) while minimizing the wirelength

Page 27: Hard Optimization Problems: Practical Approach

The hypergraph for a microchip

Page 28: Hard Optimization Problems: Practical Approach

Placement on a grid of pins

Page 29: Hard Optimization Problems: Practical Approach

Placement on a grid of pins

Page 30: Hard Optimization Problems: Practical Approach

Routing over the placement

Page 31: Hard Optimization Problems: Practical Approach

Example 3: 2D circuit placement Bottleneck in the microchip industry Given a hypergraph, find the discrete

placement of each module (gate) while minimizing the wirelength

No overlap is allowed No overflow is allowed Critical paths must be shorter Leave white space for routing

Typical IBM chip ~270 meters on 1cm2

Page 32: Hard Optimization Problems: Practical Approach

Place and route

Page 33: Hard Optimization Problems: Practical Approach

Exponential growth of transistors for Intel processors

Page 34: Hard Optimization Problems: Practical Approach

INTRODUCTION

What is an optimization problem? Examples Summary of difficulties

Page 35: Hard Optimization Problems: Practical Approach

Difficulties:

Many variables: 106 , 107 … Many constraints: 106 , 107 … Multitude of local optima Discrete nature Conflicting objectives

Reasonable running time

Page 36: Hard Optimization Problems: Practical Approach

INTRODUCTION

What is an optimization problem? Examples Summary of difficulties Is the global optimum really

needed / obtainable?

Page 37: Hard Optimization Problems: Practical Approach

PEKO=PLACEMENT EXAMPLE WITH KNOWN OPTIMUM

Place the nodes – this is the solution

Create the net list locally and compactly

The optimum wire length – the sum of all the edges between the nodes, is known and can be proven to be minimal

Page 38: Hard Optimization Problems: Practical Approach

SOLUTION QUALITY

Page 39: Hard Optimization Problems: Practical Approach

INTRODUCTION

What is an optimization problem? Examples Summary of difficulties Is the global optimum really

needed / obtainable? What is expected of a “good approximate”

solution?

Page 40: Hard Optimization Problems: Practical Approach

“Good approximate” solution

As optimal as possible:

high quality solution

Achievable in linear time

Scalable in the problem size

Page 41: Hard Optimization Problems: Practical Approach

RUNTIME

Page 42: Hard Optimization Problems: Practical Approach

Reality Check

RigorousOptimizationTheoremsLIMITED

Industrial Need for

FAST & GOOD

NP-CompleteIntractableProblems

HEURISTICS

Page 43: Hard Optimization Problems: Practical Approach

INTRODUCTION What is an optimization problem? Examples Summary of difficulties Is the global optimum really

needed / obtainable? What is expected of a “good approximate”

solution?

Multilevel philosophy

Page 44: Hard Optimization Problems: Practical Approach

MULTILEVEL APPROACH

PARTIAL DIFFERENTIAL EQUATIONS (Achi Brandt since the early 70’s) STATISTICAL PHYSICS CHEMISTRY IMAGE SEGMENTATION TOMOGRAPHY GRAPH OPTIMIZATION PROBLEMS

Page 45: Hard Optimization Problems: Practical Approach
Page 46: Hard Optimization Problems: Practical Approach

SOLUTION QUALITY

Page 47: Hard Optimization Problems: Practical Approach

SOLUTION QUALITY

Page 48: Hard Optimization Problems: Practical Approach

ORIGINAL PICTURE

Page 49: Hard Optimization Problems: Practical Approach

ORIGINAL FENG SHUI (1) FENG SHUI (2) mPL

KRAFTWERK CAPO DRAGON OURS

Page 50: Hard Optimization Problems: Practical Approach

OUR PLACEMENT

Page 51: Hard Optimization Problems: Practical Approach

Course outline 1st lecture: Introduction and motivation 2nd – 4th : Local processing (relaxation) Quadratic minimization, Newton’s method, Steepest descent, Line search,

Lagrange multipliers, Active set approach, Linear programming

4th – 5th : Global approaches Simulated annealing, Genetic algorithms, Spectral method

6th : Classical geometric multigrid 7th : Algebraic multilevel 8th : Graph coarsening 9th – 12th : Advanced multilevel topics