I. ITEMS BY A. GEOFFRION

B. INFORMAL NOTES BY A. GEOFFRION NOT INTENDED FOR GENERAL DISTRIBUTION

(in chronological order of composition)
These items are intended primarily for people who are actively engaged in structured modeling research or applications. Requests by others will be considered on an individual basis. PLEASE DO NOT CITE ANY OF THESE ITEMS WITHOUT PRIOR APPROVAL.
21. "Questions and Answers on Structured Modeling," 7 pages, 8/85.
This informal piece discusses the motivation for and possible benefits of SM. Originally written for reviewers of an NSF proposal (since funded).
 

22. "Hammer and McLeod's Semantic Database Model, Including Their Tanker Monitoring Database," Informal Note, 17 pages, 3/10/86. Revised 2/87, 8/89, 10/89.

Hammer and McLeod wrote a classic database paper in 1981 entitled "Database Description with SDM: A Semantic Database Model". These notes plow into this paper in a lot of detail. Some important philosophical differences between SDM and structured modeling come to light, and these notes are instructive for anyone interested in the powers of SM from an abstract modeling perspective.
The massive schema for Hammer and McLeod's main illustrative model is of some independent interest, as it illustrates a number of relatively uncommon SML syntax options. It appears also in "A Library of Structured Models."
 

23. "Simultaneities, Definitional Cycles, and Generic Rules with Symbolic Parameters: An Example from Macroeconomics," Informal Note, 4 pages, 3/11/86. Revised 8/87, 11/87, 2/88, 3/88.

Models that involve systems of simultaneous equations seem to violate SM's acyclicity assumption, but that is not really the case. This note clarifies the issue by way of a simple example, and it makes several other points (e.g., it illustrates the use of symbolic parameters).
 

24. "Advantages of Using a Structured Modeling System to Deliver Specialized Optimization Capability: ODS as an Example," Informal Note, 17 pages, 3/14/86. Revised 8/87, 6/88.

A structured modeling system provides the core of a turnkey system for any specific class of problems. This paper points out 7 advantages of doing this, and uses the ODS distribution planning system as an illustration. The ODS schema also appears in "A Library of Structured Models."
 

25. "The Concept of a Solver Shell: GENNET as an Example," Informal Note, 17 pages, 3/17/86. Revised 4/87, 2/88, 4/88, 8/90.

A structured modeling system like FW/SM can provide a nice shell for almost any specialized solver. The basic idea is to model the "worldview" (modeling paradigm) of the solver rather than any particular application. This paper points out 5 advantages of doing this, using the GENNET generalized network flow optimizer of Brown and McBride as an illustration and the Tanglewood problem as a detailed example. The GENNET schema also appears in "A Library of Structured Models."
 

26. "A Structured Modeling Representation of the GAMS Static Model of the Mexican Steel Industry," Informal Note, 29 pages, 3/22/86. Revised 8/7/86.

GAMS (General Algebraic Modeling System) has several objectives in common with SM. These notes compare both ways of representing one of the standard GAMS demonstration models. See also the paper by Kendrick and Krishnan. (Caveat: The schema given in this paper does not use the latest version of SML; for that, see the STEEL1 schema of "A Library of Structured Models".)
 

27. "Using Structured Modeling to Describe the Lexical and Syntactic Structure of Languages: The Case of SML," Informal Note, 4 pages, 12/2/86. Revised 7/9/87.

This note explains how SM can be used to represent modeling, programming, and other languages by representing regular expressions and Backus-Naur Form for context-free grammars. Working Paper 360 actually does this for SML.
 

28. "A Structured Modeling Approach to Text Databases," Informal Note, 8 pages, 1/26/87.

Some preliminary thoughts on how SM might be useful for dealing with text databases. A detailed example is presented for a bibliographic database.
 

29. "Modeling Categorization Hierarchies," Informal Note, 17 pages, 1/28/87. Revised 7/3/87. Addendum written 5/23/89.

Certain conceptual abstractions arise repeatedly in modeling. It is worthwhile studying how SM can represent these, as the product of such study can be used repeatedly. One commonly occurring conceptual abstraction, the categorization hierarchy, is studied in depth in this note. A detailed illustrative application (a personal filing system) is included.
The Addendum was written in response to the challenge represented by the hierarchical index set innovation of Tony H uerlimann's LPL language.
 

30. "The Model/Problem/Solver Dimension of Model Design: The Case of Project Management," Informal Note, 12 pages, 5/20/87. Revised 7/3/87, 4/19/89.

The distinction between model, problem/task, and solver is fundamental to SM. Yet sometimes it is tempting to expand the scope of a model to include the central concepts of an important problem/task, or even to include concepts belonging to what should perhaps be thought of as a solver. This paper examines such issues in the classical project management context. (Caveat: The schemas in this paper do not reflect the 11/89 changes in SML, which no longer allows /ce/ elements to be used as simple variables, so please refer to the PROMGT2 schemas of "A Library of Structured Models".)
 

31. "Structured Modeling and Definitional Systems," Informal Note, 18 pages, 5/20/87.

This note explains how SM can be viewed as formalizing a certain kind of definitional system, namely those which are acyclic, correlated, grouped, typed, hierarchical, and interpreted. This opens the door to a new class of applications and provides a useful way of thinking about all sorts of other applications. Three examples are given: 1) a small selection from a mathematics dictionary, 2) a simple gradebook, and 3) the first 50 definitions from Harary's Graph Theory, Chap. 2. The Informal Note "A Structured Model of the Theoretical Concepts of SM" gives another example. My 1989 Operations Research article incorporates some of this material as motivation.
 

32. "A Structured Model of the Theoretical Concepts of Structured Modeling," Informal Note, 10 pages, 5/27/87. Revised 8/15/88.

This note applies the approach of the Informal Note "Structured Modeling and Definitional Systems" to the definitions of the 1989 Operations Research article. The result is slightly more rigorous even than the article, and facilitates understanding the relationships among the many definitions of SM.
 

33. "Decision Analysis by Structured Modeling," Informal Note, 8 pages, 10/13/87.

This paper gives a set of rules for reformulating decision trees as structured models in a natural way. It executes these rules for a simple but instructive example. (Caveat: The schema given in this paper does not use the latest syntax of SML, so please refer to the DECISION schema of "A Library of Structured Models".)
 

34. "Hypertext and Structured Modeling," Informal Note, 14 pages, 11/87. Major revision 1/17/89.

This note discusses in some detail the strong compatibility between SM and hypertext ideas. It proposes two enhancements to FW/SM that would turn it into a "hypertext" system applicable to materials that can be cast as structured models. These enhancements also would improve considerably the ability to browse schemata. This is an area worth exploring further.
 

35. "A 'Cyclic' Model Discussed by Blanning and Muhanna," Informal Note, 10 pages, 6/5/88. Revised 8/4/88.

Acyclicity of elemental structure is a fundamental SM assumption. Naturally, whenever a model comes along that is alleged to be cyclic, it is desirable to take a close look at it. That is what this paper does. It turns out that this example is illuminating, but poses no dilemma for SM.
 

36. "A Library of Structured Models," Informal Note, 275 pages, 6/30/88. Revised 5/89, 11/89, 8/90.

This document brings together in one place about 90 structured models written in SML. The purpose is to facilitate self-study of SML and other aspects of SM, and to provide a convenient reference work for developers and researchers.
In every case there is 1) a model name and brief explanation of where the model came from, 2) a complete schema, and 3) an exhibit giving the skeletal structure of the elemental detail tables before any table joining, that is, before Step 3 of SML's Table Structuring Procedure. Table structure after joining, sample data, and references to sample data are given for selected models.
The revisions add new models, fix errors, and keep up with evolutionary improvements in SML.
 

37. "Modeling Environments: A Structured Approach," Proposal to NSF (DRMS), 29 page Project Description + many supplements, 9/88.

This proposal, which was funded, provides general support for SM research at UCLA: foundations (first order logic, conceptual modeling, extensions, scope and comparative studies), user involvement and feedback, SML evolution, and Ph.D. student projects. The Project Description makes informative reading for anyone interested in SM research.
 

38. "Structured Modeling as a Way to Describe Software Systems," Informal Note, 40 pages, 1/17/89.

This note proposes definitional systems a la SML as a style for user documentation of software systems. One example, namely the XTREE file manager for DOS, is discussed in full detail. There is a discussion of the pros and cons of this approach, both in general and by comparison with ideas put forth in a paper by Davis in Comm. ACM on software description techniques for software requirements specification.
 

39. "On Dividing SML Into Levels," Informal Note, 16 pages, 1/29/89.

SML is a complex language. It is advantageous to divide it into incrementally learnable "levels" of progressively greater expressive power. This note gives an initial four-level design that also induces a natural partition for the processes of FW/SM. The obvious advantages include good learning paths for SML and for FW/SM.
Less obvious is the possibility this suggests for a general multi-pass approach to modeling in which the modeler uses the "next" level of SML only after the model has been described -- perhaps at a relatively high level of abstraction or simplicity -- using the previous level.
This note also describes a series of possible learnability experiments in which volunteers are taught SML and FW/SM under controlled conditions. The aims would be to provide an understanding of just how hard or easy SML and FW/SM are to learn at different levels of completeness and proficiency, and to elicit guidance for the evolution of SML and FW/SM according to user needs.
See the Informal Note "Summary Notes on SML, Structured Modeling, and FW/SM". See also my January 1992 Operations Research article and Working Paper 378, which use a modified version of the 4-level partition described here.
 

40. "Computer-Based Modeling Environments," with M. Melkanoff, Proposal to NSF (CISE), 29 page Project Description + many supplements, 2/89.

This proposal sought support for the Computer Science portions of SM research. There are two main topics. The first focuses on attribute grammars and related developments for formally specifying the context-sensitive semantics of modeling languages and providing tools that can use these specifications to help automate the generation of major parts of a modeling environment. The second topic focuses on extensible database technology as an implementation platform for future modeling environments. (See the items by S. Desai and F. Vicuña in Part IIA of this Bibliography.) Many other research topics relating to SM are also summarized.
 

41. "Manufacturing Information Engineering," Proposal to IMAR (Institute for Manufacturing and Automation Research), 9 pages + appendices, 3/89.

SM is proposed as a cure for information chaos (fragmentation, redundancy, inconsistency, inscrutability) in the context of high technology manufacturing. This approach seems appropriate because it can accommodate a broad range of decision models, a rare capability among existing Information Engineering methods but a necessary one because numerous decision models need to be embedded in the control software of tomorrow's automated factory. Also, the hierarchical nature of SM is compatible with the hierarchical control system approach that is becoming the accepted standard for CIM.
The proposed project was to be carried out in three phases over three years. During the first year, we would work closely with IMAR companies in discussing the information chaos problem and possible approaches to it, carrying out a trial application of the SM approach at one or more cooperating companies, and conducting an in-depth evaluation of this approach. During the next year and a half, an IMAR Demonstration Prototype would be built for the new approach and used to carry out a more extensive demonstration application within one or more IMAR companies. The final half year would be spent refining the approach and the IMAR Demonstration Prototype, and doing technology transfer.
 

42. "Model Queries," Informal Note, 24 pages, 4/5/89. Revised 9/28/89.

The purpose of this note is to help organize thinking about "model queries" in the context of SML. What does it mean to query a model, and why is it important to be able to do so? How should one go about expressing and answering various kinds of model queries?
Numerous examples are given of three types of queries: (1) queries that can arise when one wishes to retrieve information about, or get a better understanding of, the model at hand; (2) queries that can arise when one wishes to integrate two models using the approach described in my paper "Reusing Structured Models via Model Integration"; and (3) queries that can arise when one is formulating a model or extending an existing formulation.
Three broad approaches to answering such queries are considered in detail: directly browsing the schema and elemental detail tables, using mechanically produced derivative documentation, and using a model query "language".
 

43. "Structured Modeling and Simulation," Informal Note, 32 pages, 4/24/89. Revised 10/5/89.

This note takes an example-based look at three of the main concepts characteristic of simulation: (a) random variables and stochastic processes, (b) dynamic behavior rules describing the behavior of a system over time, and (c) the notion of an experimental plan addressing system behavior over time or over repeated trials. The examples: Normal random variables and Poisson arrival processes for (a), the D/D/1 FCFS queueing system for (b), and Monte Carlo simulation of a simple structural mechanics problem, of the classical newsboy problem, and of critical path length for (c). In each case, I attempt to show the main SML modeling options. The conclusions reached are encouraging if tentative, and should be useful to anyone interested in simulation applications of SM.
(Caveat: A few of the schemas in this paper do not use the latest SML syntax. Most of the syntax changes are confined to Model 13, a current version of which appears as the Q_SIM schema in "A Library of Structured Models".)
 

44. "Maxims for Modelers," Informal Note, 25 pages, 9/10/89.

This note explains many of my beliefs about how to do successful applied work. It was written to (1) serve as a checklist to use when working on applications, (2) share with others what I have learned about applied work, and (3) clarify some of the behaviors that an ideal modeling system or environment ought to support (trial designs can then be judged in light of these). The maxims, about 20 of them, are organized into two broad groups: (A) personal conduct and project management, and (B) conceptualizing and designing models. Discussion about how SM does or does not support these maxims appears in an appendix.
 

45. "Structural Modeling with Level 1 SML, Including an Example from Date," Informal Note, 7 pages, 9/30/89.

Structural models have been discussed for years by a community of people who rely on simple graph theory to express structural relationships in an extremely wide domain of possible applications (e.g., the books by Warfield and by Harary, Norman, and Cartwright). Level 1 SML may be a good language for at least some kinds of structural modeling (Level 1 SML uses no /a/, /f/, or /t/ genera, and no indices, and so is extremely simple). Level 1 SML is also a good tool for disciplined reading and writing. This note expands on these claims, mentions connections to works of others, and gives a detailed example of Level 1 SML as a reading tool applied to some material on relational normalization theory from a well known book.
 

46. "Summary Notes on SML, Structured Modeling, and FW/SM," Informal Note, 15 pages (plus exercises and answers), 11/7/89.

These notes were prepared for a full-day SML tutorial and FW/SM demonstration given at UCLA on 11/11/89. They provide the first decent learning path for people interested in learning SML and finding out what FW/SM can do. Both SML and FW/SM are divided into 4 "levels" using a modification of the levels proposed in my note "On Dividing SML Into Levels". Working Paper 378 is a leisurely expansion of these notes into a self-contained tutorial on SML.
 

47. "Revisions to Working Paper 360 and SML," Informal Note, 10 pages, 11/7/89. Revised 11/27/89, 7/30/90.

The November 1989 and July 1990 revisions of Working Paper 360 on SML contain numerous corrections, clarifications, additions, deletions, and other alterations. This note summarizes all of the significant changes that need to be understood by a person who has already invested time in studying 360.
 

48. "Converting SML to Conventional Algebraic Notation," Informal Note, 32 pages, 11/89.

It is not difficult to translate an SML schema into algebraic notation that is easy to read and fairly conventional. One can do this by following the "rewrite rules" given in this note. Some examples are worked in detail. This note also explains how to derive the default index set of any externally indexed genus paragraph.
 

49. "Application-Specific Delivery Systems for Optimizer Engines," Informal Note, 12 pages, 10/17/90. Revised 5/27/92. See item 17 of this Bibliography.

Many worthy optimization engines are never put to practical use because it is not cost-effective to build delivery systems for them. This note suggests how this difficulty might be overcome for some optimization engines by: (a) distinguishing between "surface" and "deep" model structure, (b) using SML and a general-purpose structured modeling system to make front and back end surface structure easily tailorable to the application at hand, and (c) interfacing the engine once only for the deep structure (which would be fixed with this approach for each engine). If used to build delivery systems for several optimizer engines, this approach should yield economies of scale.
 

50. "SML Model Hierarchies," Informal Note, 10 pages + appendices, 7/29/91.

This note explores how families of SML schemas can be organized in hierarchies inspired by class hierarchies in object-oriented programming. Five criteria for judging any proposed model hierarchy are examined: simplicity, whether the hierarchy conforms to the "is a" relation, inheritance promotion, naturalness, and coverage. A 15-schema SML hierarchy is given for network flow models, and is discussed relative to the criteria.

Back to Main Page
Back to Previous Section of Annotated Bibliography
Forward to Section II of Annotated Bibliography 
Email and Web Addresses