The authors suggest that there is a way to map program trees into circuits. They start the process with a very simple embryonic electrical circuit which contains certain fixed and invariant elements for the circuit that is do be designed as well as certain wires that are capable of subsequent modification. An electirical cuicuit is progressuvely developed by applying the functions to the wires.
The evaluation of fitness for each individual circuit-constructing program tree in the population begins with its execution. The execution applies the function in the program tree to the embryonic circuit into a fully developed circuit. The authors measure fitness in terms of the sum over the fitness cases, of the absolute weighted deviation between the actual value of the output voltage (target value).
The results of the experiment showed that the resulting circuit produces an output voltage in the correct band for incoming signals emanating from the first source, the second source, or neither. It took the authors 10**15 operations (which is approximately equivalent to 1 brain second) to complete the experiment. Moreover, the performance resuts of fours different versions of genetic programming is slightly superior to that of algorithms written by knowledgeable human investigators.
The paper shows that the genetic programming is a technique which performs very well when applied to source identification problem.