Evolvability of Instruction-Based Random Boolean Networks
MetadataShow full item record
Random Boolean Networks are a generalisation of binary Cellular Automata,without a fixed topology. This thesis presents an RBN implementation usingan instruction-based approach, and compares this to a traditional table-basedapproach. The implementations are used to evolve RBNs with maximumattractor lengths, in order to investigate the evolvability and the usefulnessof an instruction-based implementation. The results show limited usefulnessfor K = 2, but the instruction-based implementation performs significantlybetter for K = 3. The instruction-based implementation is slower than thetable-based implementation by a factor of ∼ 10, but areas of improvementhave been identified and discussed.