WPC 2 B V V Z Z CG Times (Scalable) 8 w C ; , WX w P 7 X P " m ' ? ^;C dd CCCd CCCCddddddddddCC Y ~ ~w CN ~ s k~ CCCddCYdYdYCdd88d8 ddddJN8dd ddYYdY C dd dddCCCC dddddd ddd8 Y Y Y Y Y Y~Y~Y~Y~YC8C8C8C8 d d d d d d d d d d Y d d d d dsd Y Y Y Y Y Y Y d~Y~Y~Y~Y d d d d d d d dC8C8C8C8 oN d~8~8~8~8~8 d v d d d d J J JkNkNkNkN~8~8~8 d d d d d d d Y Y Y d~8 d JkN~8 d d d d d C dd C CC/ N d ddCYQQdd ddd dFdddd F CC hhd 44 ddzz d d d w oo dCh d F" d h dÕ dCC z xC d dod dCd Yds z Uw d Y Y C C C C z~o zo Y~N Y d YC8 Y o o Y d Y zsdzd d~Y Y z o zzzzNd88YYYzYz z zz CCddddd dd zzzzzzzzzzzzzzzzzzzNNNNNNNdddddddddddddddddddd888888888888YYYYYYYYYYYYYYYYYYYzzzzzzzzzzzzzzzzzzzzC s ~ C zC d dYC xHP LaserJet III HP_LJ_3.PRS o P C , , X P 2 f V V #| w 7 i C 3 , X i P 6 X P Times New Roman 2 Z HP LaserJet 4 HPLAS4.PRS o P C X P " m ^3ETgg EEEgt3E39gggggggggg99ttt ~r EP ~ r r~ ~E9E`gE g g Egg99g9 ggggEP9gg gg c)co E3EE "EEE C EEEEEE dEg9 Y Y Y Y Y Y~Y~Y~Y~YC8C8C8C8 d d d d d d d d d d Y g g d d dsd g~ ~ ~ ~ g g g g g g g gE8E9E9E9 oP g~9~9~9~8~9 g v g g g g E E ErPkNrPrP~9~9~9 g g g g g g g~ Y~ g~9 g ErP~9 g g g g gNH 3 gE gggg g9@ gFdddg F %C g EE ggzz d d d w rr E d F 9311861 Rosen NSG CGP Science Fellowship Program: Genetic Algorithms and Simulated Annealing This award will enable Dr. Bruce Rosen of the University of Texas at San Antonio to conduct collaborative research for six months with Dr. Ryohei Nakano at NTT Kyoto Research Laboratories and with Dr. Tadahiro Kitahashi at Osaka University. The team will test and analyze hybrid algorithms which Dr. Rosen developed by combining genetic algorithms and simulated annealing. Solutions to numerical problems often involve fitting a set of parameters to minimize a function. Deterministic algorithms may be used to efficiently and quickly find solutions, however, they are not guaranteed to find the optimal function state. Two relatively new methods, Genetic Algorithms (GA) and Simulated Annealing (SA), perform well on both numerical and combinatorial optimization tasks and could be combined to improve performance and runtime characteristics. Genetic Algorithms are population based search strategies which are fast and inherently parallel. Simulated Annealing algorithms employ randomness to determine new parameter values and rely on stochastic importance sampling to find the optimal function value. Rosen, Nakano and Kitahashi will investigate different methods of combining the desirable properties of both simulated annealing and genetic algorithms and test the performance of these hybrid algorithms on difficult optimization problems. Dr. Nakano's expertise in applied, real world optimization genetic algorithms and Dr. Kitahashi's expertise in simulated annealing algorithms for optimizing computer vision systems complement Dr. Rosen's experience in combining the two types of algorithms to optimize neural networks. By combining their efforts, the team will produce a fast, efficient, parallelizable hybri d GA SA algorithm for both combinatorial and numerical optimization.
