PARALLEL GENETIC ALGORITHM FOR MAGNETOTELLURIC INVERSION WITH GPU
Keywords:
Magnetotelluric inversion, Parallel genetic algorithm, CUDA architecture, Island based GAAbstract
A parallel genetic algorithm (GA) for magnetotelluric inversion with CUDA architecture is implemented for improving the accuracy and speed of traditional genetic algorithm. The algorithm is modified to adapt to the CUDA architecture for a more efficient computation. Model verification shows that the inversion computational speed has been dramatically increased with high computational accuracy under the parallel computing architecture. The CUDA architecture is proved to be a powerful tool for parallelizable problems in computational geophysics.References
[1] Eraser A S. Simulation of genetic systems by automatic digital computers. I. Introduction. Australian Journal of Biological Sciences, 1957, 10: 484-491.
[2] Holland J H. Genetic algorithms and the optimal allocation of trials. SIAM Journal on Computing, 1973, 2(2): 88-105.
[3] Everett M E, Schultz A. Two-dimensional nonlinear magnetotelluric inversion using a genetic algorithm. Journal of Geomagnetism and Geoelectricity, 1993, 45(9): 1013-1026.
[4] Perez-Flores M A, Schultz A. Application of 2-D inversion with genetic algorithms to magnetotelluric data from geothermal areas. Earth, Planets and Space, 2002, 54(5): 607-616.
[5] Roux E, Moorkamp M, Jones A G, et al. Joint inversion of long-period magnetotelluric data and surface-wave dispersion curves for anisotropic structure: Application to data from Central Germany. Geophysical Research Letters, 2011, 38(5).
[6] Wait J R. On the relation between telluric currents and the earth’s magnetic field. Geophysics, 1954, 19(2): 281-289.
[7] Wait J R. Theory of magnetotelluric fields. J. Res. NBS D, 1962, 66(5): 509-541.
[8] Pospichal P, Jaros J, Schwarz J. Parallel genetic algorithm on the cuda architecture. In Applications of Evolutionary Computation. Springer, 2010: 442-451.