Using Genetic Algorithms to Design Constant Weight Codes

Genetic algorithms have been used successfully for solving different combinatorial optimization problems. We give here a new efficient application for the search of constant weight codes. The algorithm presented is intended to generate codes with the maximum number of codewords for a given length, constant weight and minimum Hamming distance. A comparison with the simulated annealing technique is also discussed. On the other hand, the algorithm introduces a new scaling for the cost function that proved to be more effective than other scalings used in the literature. c fl1993 by Lawrence Erlbaum Assoc. Inc. Pub., Hillsdale, NJ 07642 Applications of Neural Networks to Telecommunications pp. 119--124 (1993). ISBN 0-8058-1560-0 1 Introduction The design of good codes is of fundamental importance in a communication system. Information to be transmitted is represented via a source code as the ASCII code, Huffman code, etc. This source-encoded information is then sent over a channel, such..