Submission Title

A Parallel Genetic Algorithm for Book Embedding

Presenter Information

Chauncy Cullitan, Gonzaga University

Session Number

SS1B

Location

Robinson 310

Abstract Number

SS1B-e

Abstract

Assuming a circular ordering of a graph's vertices, a book embedding is defined as an assignment of colors to the graph's edges, such that no same-colored edges intersect. The minimum number of colors required to produce a valid book embedding is the graph's book thickness. Computing the book thickness of an arbitrary graph is generally intractable by brute force means, so we use genetic algorithms to find approximate solutions. A genetic algorithm (GA) walks through the search space with a population of candidate solutions, in a manner loosely inspired by biological evolution.

We present a simple GA (written in C++) for approximating the optimum book thickness of a graph with fixed vertex ordering. We also give our preliminary results with a massively parallel GA on the CUDA architecture, which is approximately 3x faster than the single-threaded GA.

This document is currently not available here.

COinS
 
Apr 23rd, 10:45 AM Apr 23rd, 12:15 PM

A Parallel Genetic Algorithm for Book Embedding

Robinson 310

Assuming a circular ordering of a graph's vertices, a book embedding is defined as an assignment of colors to the graph's edges, such that no same-colored edges intersect. The minimum number of colors required to produce a valid book embedding is the graph's book thickness. Computing the book thickness of an arbitrary graph is generally intractable by brute force means, so we use genetic algorithms to find approximate solutions. A genetic algorithm (GA) walks through the search space with a population of candidate solutions, in a manner loosely inspired by biological evolution.

We present a simple GA (written in C++) for approximating the optimum book thickness of a graph with fixed vertex ordering. We also give our preliminary results with a massively parallel GA on the CUDA architecture, which is approximately 3x faster than the single-threaded GA.