AXIOMATIC DESIGN OF ALGORITHMS FOR THE AUTOMATION OF COMPLEX REAL-WORLD TASKS

We propose an approach to the design of algorithms for the automation of complex real-world tasks. The approach is based on the view that th e key to success lies in the ability, to reduce computational complexi ty. In general, knowledge of the task is a critical factor in enabling effective reduction of computational complexity, but knowledge of pro cedural matters is also important and helpful. We address the latter i ssue ill this paper proposing some axioms to guide the complexity redu ction procedure. Some enabling technologies include pattern recognitio n, neural-net computing, and evolutionary programming. We argue that e xpertise in the use of such tools would not only facilitate the exerci se of the axioms but would also help ensure that the algorithm remains adaptive in both the acquisition and use of knowledge of the task.