Toward a comprehensive, knowledge-based software design course

Many colleges and universities teach courses in the design and implementation of so-called expert systems. Such a course might include discussions of topics from traditional (symbolic) artificial intelligence, and culminate in the construction of a prototypical rule-based system. In general, the emphasis is on the use of knowledge compiled and encoded by the designer. Here, an alternative approach is suggested in which such a course is broadened to include fundamental topics from the study of artificial neural nets. In this way the student is exposed to the notion that not only can program knowledge be contributed directly by the designers, but might also be extracted by the software from the data itself.