Developing a generic genetic algorithm

Genetic and evolutionary algorithms, inspired by biological processes, provide a technique for programs to "automatically" improve their parameters. We discuss the basics of the algorithms and introduce our own hybrid. The development of this hybrid and its application to a simplified problem, evolving the coefficients for the sine function in a Taylor series, presents opportunities for computer science education with respect to model-building, data structures, and language features. Students must decide upon the representation of the chief mechanisms of genetic algorithms: mutation to alter the values of parameters directly and crossover to vary the groupings of co-evolved parameters in order to break away from local fitness maxima. They must examine the meaning of fitness itself as well as make many other modeling decisions. Ada itself provides both challenges and advantages: linked-lists must be well understood to be updated in an object-oriented context and hard-typing produces mixed reactions in students used to C++, but generics provide a powerful way to generalize the algorithm and incorporating different problem domains.