COMP 5704

Parallel Algorithms and Applications in Bioinformatics

 

Frank Dehne, PhD (www.dehne.net)
Professor of Computer Science

Carleton University
Ottawa, Canada

Parallel Computing Project
Name: Dave McKenney
URL: http://people.scs.carleton.ca/~dmckenne/5704/front_page.htm
E-mail: dmckenne at connect dot carleton dot ca

Project Title: Genetic Programming on a GPU Applied to the Creation of Stock Trading Rules

Project Outline:

This project has several goals. The first goal is to implement a genetic programming (GP) approach to creating stock trading rules. This will most likely be done using an already developed C genetic programming tool called lilgp. Second, this GP will be modified to allow evaluation of individuals within the population on a nVidia GPU device using CUDA. Realization of this second step will involve both the modification of the general GP functionality, as well as the implementation of an interpreter which takes GP individuals as data and evaluates there fitness on a GPU device. Finally, several approaches to evaluating individuals on a GPU will be tested against each other. This includes both the BlockGP and ThreadGP population parallel methods as outlined in (need reference), as well as a fitness parallel approach in which one individual is evaluated at a time on the GPU (with fitness cases being run in parallel).

Startup Paper(s):

[1]A Data Parralel Approach to Genetic Programming Using Programmable Graphics Hardware
[2]An Implementation of Genetic Algorithms as a Basis for a Trading System on the Foreign Exchange Market
[3]High Performance Genetic Programming on GPU
[4]A Comparison of Genotype Representations to Acquire Stock Trading Strategy Using Genetic Algorithms

Deliverables:

Relevant References: