EA* A Hybrid Approach Robbie Hanson. What is it?  The A* algorithm, using an EA for the heuristic.  An efficient way of partitioning the search space.

Slides:



Advertisements
Similar presentations
Review: Search problem formulation
Advertisements

Hard Problems Some problems are hard to solve. No polynomial time algorithm is known Most combinatorial optimization problems are hard Popular NP-hard.
State Space Representation and Search
Traveling Salesperson Problem
CS6800 Advanced Theory of Computation
Heuristics CPSC 386 Artificial Intelligence Ellen Walker Hiram College.
Part2 AI as Representation and Search
Artificial Intelligence Lecture
Anytime RRTs Dave Fergusson and Antony Stentz. RRT – Rapidly Exploring Random Trees Good at complex configuration spaces Efficient at providing “feasible”
Institute of Intelligent Power Electronics – IPE Page1 Introduction to Basics of Genetic Algorithms Docent Xiao-Zhi Gao Department of Electrical Engineering.
1 Wendy Williams Metaheuristic Algorithms Genetic Algorithms: A Tutorial “Genetic Algorithms are good at taking large, potentially huge search spaces and.
Hybridization of Search Meta-Heuristics Bob Buehler.
Branch and Bound Similar to backtracking in generating a search tree and looking for one or more solutions Different in that the “objective” is constrained.
MAE 552 – Heuristic Optimization Lecture 27 April 3, 2002
MAE 552 – Heuristic Optimization Lecture 26 April 1, 2002 Topic:Branch and Bound.
Imagine that I am in a good mood Imagine that I am going to give you some money ! In particular I am going to give you z dollars, after you tell me the.
Problem Solving and Search in AI Heuristic Search
5-Nov-2003 Heuristic Search Techniques What do you do when the search space is very large or infinite? We’ll study three more AI search algorithms: Backtracking.
Ant Colony Optimization Algorithms for the Traveling Salesman Problem ACO Kristie Simpson EE536: Advanced Artificial Intelligence Montana State.
Backtracking.
Informed Search Idea: be smart about what paths to try.
Genetic Algorithms: A Tutorial
Genetic Algorithm.
Programming & Data Structures
Genetic Algorithms and Ant Colony Optimisation
Computer Implementation of Genetic Algorithm
Branch & Bound UPPER =  LOWER = 0.
1 Paper Review for ENGG6140 Memetic Algorithms By: Jin Zeng Shaun Wang School of Engineering University of Guelph Mar. 18, 2002.
A Genetic Solution to the Travelling Salesman Problem Ryan Honig.
CSCE350 Algorithms and Data Structure Lecture 17 Jianjun Hu Department of Computer Science and Engineering University of South Carolina
CS440 Computer Science Seminar Introduction to Evolutionary Computing.
A Genetic Solution to the Travelling Salesman Problem Ryan Honig.
Optimization Problems - Optimization: In the real world, there are many problems (e.g. Traveling Salesman Problem, Playing Chess ) that have numerous possible.
1 “Genetic Algorithms are good at taking large, potentially huge search spaces and navigating them, looking for optimal combinations of things, solutions.
Applications of Dynamic Programming and Heuristics to the Traveling Salesman Problem ERIC SALMON & JOSEPH SEWELL.
For Friday Finish reading chapter 4 Homework: –Lisp handout 4.
For Monday Read chapter 4, section 1 No homework..
CSE 589 Part VI. Reading Skiena, Sections 5.5 and 6.8 CLR, chapter 37.
1 Genetic Algorithms and Ant Colony Optimisation.
1 Branch and Bound Searching Strategies Updated: 12/27/2010.
Introduction to Genetic Algorithms. Genetic Algorithms We’ve covered enough material that we can write programs that use genetic algorithms! –More advanced.
For Wednesday Read chapter 6, sections 1-3 Homework: –Chapter 4, exercise 1.
Slides by: Eric Ringger, adapted from slides by Stuart Russell of UC Berkeley. CS 312: Algorithm Design & Analysis Lecture #36: Best-first State- space.
Ricochet Robots Mitch Powell Daniel Tilgner. Abstract Ricochet robots is a board game created in Germany in A player is given 30 seconds to find.
Analysis of the Traveling Salesman Problem and current approaches for solving it. Rishi B. Jethwa and Mayank Agarwal. CSE Department. University of Texas.
Chapter 3.5 and 3.6 Heuristic Search Continued. Review:Learning Objectives Heuristic search strategies –Best-first search –A* algorithm Heuristic functions.
Gerstner Lab, CTU Prague 1Motivation Typically,  an evolutionary optimisation framework considers the EA to be used to evolve a population of candidate.
Genetic Algorithms and TSP Thomas Jefferson Computer Research Project by Karl Leswing.
Metaheuristics for the New Millennium Bruce L. Golden RH Smith School of Business University of Maryland by Presented at the University of Iowa, March.
Best-first search is a search algorithm which explores a graph by expanding the most promising node chosen according to a specified rule.
Artificial Intelligence By Mr. Ejaz CIIT Sahiwal Evolutionary Computation.
For Monday Read chapter 4 exercise 1 No homework.
Genetic Algorithms. Solution Search in Problem Space.
Chap 4: Searching Techniques Artificial Intelligence Dr.Hassan Al-Tarawneh.
Paper Review for ENGG6140 Memetic Algorithms
Comparing Genetic Algorithm and Guided Local Search Methods
Chap 4: Searching Techniques Artificial Intelligence Dr.Hassan Al-Tarawneh.
metaheuristic methods and their applications
Genetic Algorithms: A Tutorial
Branch and Bound.
Heuristic Algorithms via VBA
Metaheuristic methods and their applications. Optimization Problems Strategies for Solving NP-hard Optimization Problems What is a Metaheuristic Method?
Blay Whitby 2003 Search Blay Whitby 2003
Backtracking and Branch-and-Bound
Chap 4: Searching Techniques
Heuristic Algorithms via VBA
Heuristic Algorithms via VBA
More advanced aspects of search
Traveling Salesman Problem by Genetic Algorithm
Genetic Algorithms: A Tutorial
Presentation transcript:

EA* A Hybrid Approach Robbie Hanson

What is it?  The A* algorithm, using an EA for the heuristic.  An efficient way of partitioning the search space for an EA.  An effective termination technique for EA’s.  The A* algorithm, using an EA for the heuristic.  An efficient way of partitioning the search space for an EA.  An effective termination technique for EA’s.

Motivation  Poor EA performance  Local maxima/minima traps!  Balancing exploration with exploitation.  When to terminate?  Poor EA performance  Local maxima/minima traps!  Balancing exploration with exploitation.  When to terminate?

The General Idea  Partition the search space  Explore each partition  Continue exploration on “promising” partitions  Partition the search space  Explore each partition  Continue exploration on “promising” partitions

Motivation (cont)  Hybridisation… (Chapter 10)  “This category of algorithms is very successful in practice and forms a rapidly growing research area with great potential.”  Hybridisation… (Chapter 10)  “This category of algorithms is very successful in practice and forms a rapidly growing research area with great potential.”

Intro to A*  Branch and bound technique  Extension of best-first search of a tree  Uses heuristic to determine fitness of nodes  Branch and bound technique  Extension of best-first search of a tree  Uses heuristic to determine fitness of nodes

Example Tree

Problems with A*  A* relies on a good heuristic  Without it, it becomes essentially a breadth first search  Some problems are difficult to design standard heuristics for  “For every common sense heuristic you can invent, you can find a pathological case that will make it look very silly.”  (Michalewicz & Fogel: “How to solve it: Modern Heuristics”)  A* relies on a good heuristic  Without it, it becomes essentially a breadth first search  Some problems are difficult to design standard heuristics for  “For every common sense heuristic you can invent, you can find a pathological case that will make it look very silly.”  (Michalewicz & Fogel: “How to solve it: Modern Heuristics”)

Traveling Salesman Problem The traveling salesman must visit every city in his territory exactly once and then return home covering the shortest distance. Search space: (N-1)! / 2 10-city: 181,000 solutions 20-city: 10,000,000,000,000,000 solutions TSPlib contains many real world examples The traveling salesman must visit every city in his territory exactly once and then return home covering the shortest distance. Search space: (N-1)! / 2 10-city: 181,000 solutions 20-city: 10,000,000,000,000,000 solutions TSPlib contains many real world examples

Example

EA Details  Parameter file specifies specifics, such as population size, number of children, etc.  Log file captures output.  This facilitates experimentation of parameter values.  Parameter file specifies specifics, such as population size, number of children, etc.  Log file captures output.  This facilitates experimentation of parameter values.

Representation and Fitness

Selection and Survival  Tournament Selection  Selection size specified in parameter file  (µ + λ) survival strategy  Tournament Selection  Selection size specified in parameter file  (µ + λ) survival strategy

Recombination/Mutation  Single parent mutation most popular  Two popular methods  Single parent mutation most popular  Two popular methods

EA* specific  Number of generations to run EA for each iteration.  How long may a node remain in the “open list?”  Number of generations to run EA for each iteration.  How long may a node remain in the “open list?”

Performance  Final solutions are VERY consistent.  Initial results suggest a lower standard deviation than regular EA.  SO FAR, it averages better solutions. (Very difficult to say)  Final solutions are VERY consistent.  Initial results suggest a lower standard deviation than regular EA.  SO FAR, it averages better solutions. (Very difficult to say)

Problems  Large TSP problems  Lucky first guesses  Large TSP problems  Lucky first guesses

Future Research  EA’s report expected fitness in generations to come.  This could help the EA to overestimate less often, possibly making the heuristic admissible for A*.  Local search techniques in the EA for better performance.  Trivial parallelization. (BOINC?)  EA’s report expected fitness in generations to come.  This could help the EA to overestimate less often, possibly making the heuristic admissible for A*.  Local search techniques in the EA for better performance.  Trivial parallelization. (BOINC?)

Questions?