Keith Robert Milliken

A Ramsey theorem for trees

Abstract We prove a Ramsey theorem for trees. The infinite version of this theorem can be stated: if T is a rooted tree of infinite height with each node of T having at least one but finitely many immediate successors, if n is a positive integer, and if the collection of all strongly embedded, height- n subtrees of T is partitioned into finitely many classes, then there must exist a strongly embedded subtree S of T with S having infinite height and with all the strongly embedded, height- n subtrees of S in the same class.

A continuous real-time expert system for computer operations

The Yorktown Expert System/MVS Manager (or YES/MVS for short) is a continuous real-time expert system that exerts active control over a computing system and provides advice to computer operators. YES/MVS provides advice on routine operations and detects, diagnoses, and responds to problems in the computer operators domain. This paper discusses the YES/MVS system, its domain of application, and issues that arise in the design and development of an expert system that runs continuously in real time.

YES/MVS and the automation of operations for large computer complexes

The Yorktown Expert System/MVS Manager (known as YES/MVS) is an experimental expert system that assists with the operation of a large computer complex. The first version of YES/MVS (called YES/MVS I) was used regularly in the computing center of IBMs Thomas J. Watson Research Center for most of a year. Based on the experience gained in developing and using YES/MVS I, a second version (YES/MVS II) is being developed for further experimentation. This paper discusses characteristics of the domain of large computing system operation that have been illuminated by the YES/MVS I experience, and it describes the modifications in the design of YES/MVS II that are an outgrowth of the YES/MVS I experience.

Adding rule-based techniques to procedural languages

The world of software for applications in industry is dominated by compiled, algebraic (Algol derivative) programming languages such as C, Pascal, and PL/I. One approach to popularizing expert systems techniques in the industrial environment is to integrate software techniques appropriate for expert systems development into algebraic programming languages. While programmers in industry seek more powerful software tools, generally they do not want to give up existing capabilities in order to gain access to new techniques. Thus the question becomes how to integrate expert system techniques into a procedural language so that important characteristics of the procedural language are preserved. YES/LI (Yorktown Expert System / Language One) is an experimental tool for developing expert systems that is an integration of PL/I with data-driven rules. This paper reviews a number of technical issues encountered in the development of YES/LI. The achieved resolution of these issues establishes that rule-based techniques intended for use in large, industrial expert systems can be integrated into the standard features of an existing, algebraic programming language without reducing the functionality of the underlying, procedural language. Features of YES/LI include: condition-action rules with support for both highly expressive conditions and complex actions, PL/I as a subset, block structuring including recursively invoked blocks of rules, support for large blocks of rules written in several source code files, the same access to data bases and routines written in other languages as found in PL/I, excellent performance, portability across several operating systems and computer architectures, the ability for YES/LI routines to be linked, loaded, started, and called just as PL/I routines.