FSM CONSTRUCT WITH GENETIC ALGORITHM USING DISTANCE MEASURING IN MUTATION OPERATOR

Our approach to the FSM construction using genetic algorithm was presented in a several contributions. The FSM was represented by matrices. The one-point crossover operator mixed two matrices and the mutation changed one transition or output. Very difficult tasks were these ones where FSMs recognize given subsequences in the serial bit sequences. Transition graphs of these FSMs contained trees as their sub-graphs. The algorithm achieved certain fitness and then it stagnated on. We modified the mutation operator. We defined and computed the distance between two FSMs and we add the distance measurement into the mutation operator. Now the mutation changes the FSM so that the distance between the original FSM and mutated one is greater than predefined threshold.