EASEA parallelization of tree-based Genetic Programming
Image Sci., Comput. Sci. & Remote Sensing Lab., Univ. of Strasbourg, Illkirch, France
IEEE Congress on Evolutionary Computation (CEC), 2010
This paper introduces the implementation of Koza-style tree-based Genetic Programming on General Purpose Graphic Processing Units (GPGPU) using the EASEA language, and shows how a GP algorithm can be easily implemented using EASEA and CUDA. Performance is first discussed on a classical toy problem taken from one of Koza’s books and then on a real world problem inspired from aeronautics, that extends the results to difficult problems with large data sets.
January 22, 2011 by hgpu