A Comparative Evolutionary Models for Solving Sudoku

Abstract

Evolutionary algorithms have become robust tool in data processing and modeling of dynamic, complex and non-linear processes due to their flexible mathematical structure to yield optimal results even with imprecise, ambiguity and noise at its input. The study investigates evolutionary algorithms for solving Sudoku task. Various hybrids are presented here as veritable algorithm for computing dynamic and discrete states in multipoint search in CSPs optimization with application areas to include image and video analysis, communication and network design/reconstruction, control, OS resource allocation and scheduling, multiprocessor load balancing, parallel processing, medicine, finance, security and military, fault diagnosis/recovery, cloud and clustering computing to mention a few. Solution space representation and fitness functions (as common to all algorithms) were discussed. For support and confidence model adopted 1=0.2 and 2=0.8 respectively yields better convergence rates – as other suggested value combinations led to either a slower or non-convergence. CGA found an optimal solution in 32 seconds after 188 iterations in 25runs; while GSAGA found its optimal solution in 18seconds after 402 iterations with a fitness progression achieved in 25runs and consequently, GASA found an optimal solution 2.112seconds after 391 iterations with fitness progression after 25runs respectively.