GA with Modifications for Efficiently Solving Number Puzzles

Anagha P. Khedkar

Abstract

The previous research on the genetic algorithm (GA) applications especially for puzzle solving focuses on the GA capability for solving the puzzles of differing difficulty levels viz. magic square, jigsaw, sliding tile, crossword, path finding and of similar type. But this work attempts to solve logical number puzzles of exponential and factorial complexity using conventional GA with some important modifications depending on the puzzle type. The overall modifications done in GA are the advanced twin operator, order crossover and use of only mutation without crossover. GA is designed and applied with these modifications considering one at a time to solve the three number puzzles. The experimental results reflect the performance improvement in GA in terms of best and average convergence generation. The advanced twin operator enhances the overall performance of modified GA by 40% over the conventional GA. The modified GA is also tested for variation in population size which shows the faster convergence with sufficiently increased population size. Further to extend the current work, it is needed to find out the optimal population size depending on the puzzle problem.Cite this ArticleKhedkar Anagha P. GA with modifications for efficiently solving number puzzles. Journal of Artificial Intelligence Research & Advances. 2016; 3(1): 1–7p.

Abstract

Keywords