Solving large scale problems using estimation distribution algorithm with arithmetic coding

This work proposes an algorithm which combines estimation distribution algorithm with a chromosome compression scheme to solve large scale problems. The search space reduction resulted from chromosome compression enables the proposed algorithm to solve one-million-bit problems and a one-billion-bit problem. Arithmetic coding represents a compressed binary string with two real numbers. Using this representation, a model of highly fit individuals can be constructed. This model can be used to evolve the solution in the manner of estimation distribution algorithm. The proposed algorithm is applied to large scale problems which are one-million-bit OneMax, royal road, trap functions. It is also applied to one-billion-bit OneMax problem. The experimental result shows that the proposed algorithm can solve million-bit OneMax problem in 4 seconds and billion-bit OneMax problem in 92 minutes using a normal PC-class computer.