Search

This grant led to developments in flexible models for complex time series in a range of applications with a focus on Bayesian and Bayesian nonparametric methods. Three fundamental challenges were tackled: (i) capturing evolving correlations in high-dimensional time series with possible missing or irregularly-spaced observations, (ii) performing diverse subset selection over time, and (iii) automatically learning an unknown set of simple underlying temporal structures to describe complex dynamical phenomena. Each of these methods was applied in a range of application domains including neuroimaging, diverse document selection, speaker diarization, stock modeling, and target tracking.

Catastrophe insurance helps spread risks and increases the ability of policyholders and the economy to recover from both natural disasters and terrorist attacks. Government policies, however, may unintentionally limit the role of the private sector in insuring against catastrophic losses. Several such policies at both the state and the federal level reduce the amount of private capital supplied to insure or hedge against catastrophic risks. One reason is that those policies often become outdated as markets innovate. Policymakers have several different options to increase private risk-bearing capacity and improve the effectiveness of federal involvement. The benefits and potential costs of four options are examined: an optional federal charter for insurers that would preempt states regulation of rates; regulatory reform of capital markets risk transfer mechanisms that substitute for reinsurance; changes in the taxation of reserves held by insurers against catastrophic risks; and auctions of federal reinsurance for supercatastrophic risks.

These viewgraphs concern health care-economic analysis, consumers, providers, etc. in the United States.

Money spent on Weapon systems always has another use alternative. The United States procures weapons system to provide for National Security, from which every citizen benefits. All weapons systems have affordability choices. Weapon system affordability always conjures up an old debate of 'cost effectiveness at the margin.' Defense, at the lowest cost, always applies to all DOD acquisition and technology programs. From DoD's perspective, the ballistic missile defense network has to deal with the offensive weapon's major advantage of surprise. The question of how few ABM defensive locations are necessary to defeat the adversary's threat, is an affordability tradeoff issue balanced by engineering and system technology capabilities. These tradeoffs stress the DOD Acquisition Reform initiative's underlying impetus on affordability. NMD Program Management implemented DOD acquisition reform initiatives and strategies to manage for acquisition excellence and program balance during NMD's capability development and deployment planning phase. This includes measurable (affordable) characteristics of key enabling technologies for cost breaking improvements in production, procurement, support, and operations. The DOD Program Manager should have latitude for initiatives in the way the defense capability is built so that it appears as a cost-effective and affordable technology solution to a given threat. There is no affordable solution to developing an ABM defense capability unless technology can develop one., Prepared in collaboration with SAIC, CSF, Arlington, VA.

This grant has the central aim of exploring when, and how, probability is useful in the theory of algorithms. Most of the problems reviewed have their origins in the area of Euclidean Combinatorial Optimization, which might be operationally defined as the theory that has evolved out of the study Euclidean traveling salesman problem (TSP), the minimal spanning tree problem, and the minimal matching problem. Probability enters the study of such problems by two different paths. One path calls on exogenous randomization in the course of a genuine probabilistic algorithm. This path is of increasing importance in many areas, and on an elementary level is well illustrated by the method of simulated annealing. A second path of considerable importance calls on the introduction of stochastic models for the problem inputs. One then uses probability theory to understand as deeply as possible the behavior of the associated objective functions. This understanding is used subsequently to guide algorithm design.

A massively parallel algorithm for nonlinear multicommodity flow problems has been designed and tested. A new promising algorithm for quadratic stochastic programs with network structures has also been developed.

We consider a common variant of the vehicle routing problem in which a vehicle fleet delivers products stored at a central depot to satisfy customer orders. Each vehicle has a fixed capacity, and each order uses a fixed portion of vehicle capacity. The routing decision involves determining which of the demands will be satisfied by each vehicle and what route each vehicle will follow in servicing its assigned demand in order to minimize total delivery cost. We present a heuristic for this problem in which an assignment of customers to vehicles is obtained by solving a generalized assignment problem with an objective function that approximates delivery cost. This heuristic has many attractive features. It has outperformed the best existing heuristics on a sample of standard test problems. It will always find a feasible solution if one exists, something no other existing heuristic can guarantee. It can be easily adapted to accommodate many additional problem complexities. By parametrically varying the number of vehicles in the fleet, our method can be used to optimally solve the problem of finding the minimum size fleet that can feasibly service the specified demand.

During the past nine months, the focus has been on advanced computer based techniques to support decision aiding systems. This has been centered on the development and implementation of a prototype decision aiding information system, DAISY, which should aid in gaining experience with these types of systems. In January 1975, a team of visitors from the Navy and other contractors on the ODA project visited the University of Pennsylvania for a briefing on the overall plans for DAISY, and for a look at certain advanced techniques (e.g., Natural English Language input) which might prove useful in operational decision aiding systems.

This report is Volume 1 of a large scale field marketing experiment conducted over a three year period. This research was designed to measure and quantify, where possible, the effectiveness of Navy recruiting resources. The background of this research effort is presented in this report, together with a discussion of the experimental methodology and of the choice of measured endogenous and exogenous variables. This is followed by a short description of the collected data, and of the measurement techniques employed. Observed responses to experimental and environmental variables are briefly presented. This leads to an identification of the factors which affect military enlistments, and to an estimation of the magnitude of their effects. The effect of key marketing variables over time is examined. There follows a summary of supporting data obtained through tracking studies of perceptions, attitudes, and demographics. A conclusion discusses observed marginal costs and effects of the various treatments, and suggests implications for future resource allocation., See also Volume 4, ADA120504.

This report is Volume 3 of a large scale field marketing experiment conducted over a three year period. This research was designed to measure and quantify where possible the effectiveness of Navy recruiting resources. A discussion of the problems and issues of salesforce productivity measurement begins this report. After presentation of the data on which the investigation is based, observed learning and de-learning effects are described. Other significant phenomena are also discussed, among them the effects of recruiting goals, differences between regions and involuntary extensions of recruiters duty tours. The observed frequency distribution of recruiter productivity is presented. This is observed frequency distribution of recruiter productivity is presented. This is followed by a discussion of recruiter performance forecasting, and by suggestions for further research., See also Volume 5, ADA120505.

This report is Volume 7 of a large scale field marketing experiment conducted over a three year period. This research was designed to measure and quantify where possible the effectiveness of Navy recruiting resources. Sophisticated multivariate cluster analysis has been applied to the collected attitudinal data to determine the nature and size of any identifiable market segments in the at-large population of young people. This report outlines the technique and results of the study, then evaluates the differential rates at which the observed segments proceed through the Navy recruiting process. Differences which may be associated with variations in experimental treatment conditions are also identified., See also Volume 1, ADA120503.

This report is Volume 5 of a large scale field marketing experiment conducted over a three year period. This research was designed to measure and quantify where possible the effectiveness of Navy recruiting resources. Demographic, attitudinal, and perceptual data are presented in this report for the at-large population of young people, as sampled by telephone survey; and for participants in the recruiting cycle itself, as sampled through written questionnaires. A baseline is thus established for understanding of further studies. The cross-sectional view of the recruiting process leads to insights into its mechanisms. Complete tabulations of the collected data are appended., See also Volume 7, ADA120506.

The sense of time is implicit in almost all human activity, yet it is rarely reflected in the computer database views of these activities. This paper offers a method of dealing with time, modal storage and retrieval, and describes formal and practical realizations of the concept. The conceptual framework of Modal Data Management reflects three years of experience with the DATA (Dynamic Alerting Transaction Analysis) system, in which a time oriented DBMS has been conceived, designed and actually implemented. The paper first anchors the Modal Data Management concept in the context of the relevant research in Information Modelling and storage technologies, followed by a discussion of the architectural and functional attributes of the DATA system. The major lessons from the DATA experience are then presented, paying special emphasis to their impact on the design of future versions of the system. The concluding comments deal with specific implications of the Modal Data Management in the domains of DBMS usability and Software Engineering. (Author)

The flat-file alerter system allows one to monitor a flat file for various conditions of interest. A flat file is a sequential file of fixed-length records of identical format. In addition, the file is keyed on a specific field, and sorted on that key. Thus the addition of records (appearance of records with keys that did not exist before), deletion of records (disappearance of keys that existed before), and modification of records (modification of a field other than the key in a record) can be surveyed. (Author)

A method is described for integrating two or more databases into a conceptual superview, through a set of schema transformations. Such integration may be useful when it is required to produce a unified view of two databases while preserving their physical independence. Each transformation defines a mapping of queries against the superview into the appropriate set of queries against the underlying data bases and imposes a constraint that is checked when the query is evaluated. A program that interactively aids the user in constructing the superview and that performs this query transformation is being developed. (Author)

This paper addresses the problems in providing graphic displays automatically to serve a user naive with respect to computer graphic devices. It identifies the properties of data that affect graphic representation and presents a formalism in which to view them. It also discusses and illustrates the selection of various graphic formats based on the data to be represented, its properties, and graphic device characteristics. (Author)

Much of the work on relational databases that deals with data dependencies makes a uniqueness (or universal relation) assumption. It has been recognized that this assumption is problematic; nevertheless it is necessary for the axiomatic approach taken in many papers on the theory of relational databases. We will describe the problem, investigate some of the solutions put forward and suggest a new solution. Many of the problems remain intractable within the realm of 'classical' relational databases and restrictions must be placed on the use of FDs. An automated method is presented that searches for violations of the uniqueness assumption. (Author)

Structured query languages, such as those available for relational databases, are becoming increasingly desirable for all database management systems. Such languages are applicative: there is no need for an assignment or update statement. A new technique is described that allows for the implementation of applicative query languages against most commonly used database systems.

This paper presents an architecture for a generalized model management system that facilitates the integration of management science models into a decision support system. The model management system will provide a methodology for making mapping more scientific. The linking together of models with data and solution processes is a difficult and time-consuming process for a user of such models. The model management system will automate this linking as much as possible. (Author)

This paper describes a system designed to extend the traditional DSS concept in order to support the model management requirements of complex organizations. Its objectives are to provide a mechanism to represent and to diffuse the organizational knowledge about models so that the user community can utilize this knowledge to adapt or build decision aids. (Author)

A Codasyl interface has been constructed for Pascal and designed for Ada that exploits the data type systems of these languages. We believe that the form of this interface will simplify the writing of Codasyl applications and greatly reduce errors in coding. In particular, it relieves the user from the need to consider data currency and global data areas and uses the host language's type system to perform many checks at compile-time that in other interfaces can at best be done at run-time. (Author)

Despite the enormous advances that have been made in the specification of data types and data models in the fields of programming languages, databases and artificial intelligence; there remain a number of problems in attempting to unify the various approaches to the formal description of data. The purpose of this brief paper is to examine these problems from the point (or points) of view of those people -- designers, administrators, applications programmers, and end-users -- whose main interest is with databases. In particular, we hope to display special concern for the tools provided for the end-user, who should be the final beneficiary of whatever advances are made.

A satisfactory user interface is essential for a decision support system to gain acceptance, and system performance is an important component of the user interface. Several suggestions for data base design are presented to improve system performance and response time, and thus to improve the user interface. (Author)

The Window Management System (WINDOWS) is designed to overcome certain limitations on computer decision aids which stem from the fact that the user of such aids is typically restricted to interacting with one computer program at a time. The decision maker has access to a variety of different sources of information, such as books, graphs, maps, and notes. WINDOWS makes it possible to divide a single computer terminal into multiple 'sub-terminals,' thus enabling decision makers to interact with more than one computer process at a time from a single terminal. In addition, WINDOWS allows the decision maker to control all of the computer processes from a single keyboard. (Author)

This report is the Final Report for the Research Project N00014-77-C-0171 on 'Internal Financial Incentives in Systems Acquistion'. It includes as an Appendix a final version of the major theoretical output of this research: a theory of multistage contractual incentives in the systems acquisition cycle. The theory proposed links up incentives in the development and production stages of systems acquisition in a unified framework. The impact of such incentives under various government contracting policies, environment conditions and market structures is analyzed. The results are illustrated with an extended numerical example. (Author)

This paper has two central themes: the comparative analysis of alternative institutional forms of labor influence in monopolies; and the design of regulatory mechanisms for improving the welfare efficiency of such firms. Concerning the comparative analysis of different firm types, labor-managed and codetermined firms have two basic problems: underemployment and perverse short-run adjustment to price changes of output and long-run inputs. In addition, labor-managed firms may undervalue investment projects and retained earnings relative to efficient market levels. On the positive side, being the partial recipient of residual profits in the firm improves labor productivity. The policy implications of this analysis are simply stated. If overemployment or low productivity are central problems, then increasing labor's influence in the firm, with concommitant incentives, is a possible cure. Just how well this cure is likely to work relative to a traditional capitalist-managed firm depends, inter alia, on the income elasticity of workers' effort (how well they will respond to productivity incentives) and the price elasticity of demand (how strongly the market reinforces increases in output resulting from improved productivity). Regarding regulation, lump-sum taxes, and income and price regulation can improve the efficiency of labor-managed and codetermined firms. More direct regulation is specifying acceptable levels of capital input and employment will naturally also improve efficiency, gross of the transaction costs involved.

True automation in the office comes when the initiation of activities is aided by the system rather than the system merely serving as a passive tool to assist when the person asks. In this vein, it has been built upon the SCOOP and SEED database systems an office automation system which keeps track of the promises made to people, and vice versa.

As an extension to the CODASYL Data Description Language, a new external schema description language to provide an external schema facility more powerful than that presently offered by CODASYL subschemas is proposed a powerful external schema facility that, in effect, offers applications programmers a virtual data base. Programmers will use records formatted, not like those stored in the common data base, but like the cognitive structures employed in program development. A possible language for external schema definition is introduced.

If the much touted 'office of the future' is finally here, what comes next? This paper describes some of the thoughts and observations of what the research community, and more advanced users are working on and looking forward to as office automation technology becomes widespread and mature. (Author)

The CODASYL DDLC 1978 Report incorporates numerous enhancements and language changes. Unfortunately, the major design limitations associated with earlier reports and specifications, in particular a schema facility too closely related to machine rather than enterprise requirements and an extremely limited subschema facility, are retained. We suggest that the recent CODASYL specifications remain inappropriate as either an instance of an ANSI/SPARC three-schema architecture or as a candidate for a national data base system standard. (Author)

This report describes a concept of 'corporate memory' which attempts to provide information systems support to the organization as a whole. It also examines the propositions that organizations incur higher information gathering costs than individuals and higher lost opportunity costs. The corporate memory system is a logical extension to the office automation concepts being widely discussed today. A proposed design and some examples of potential utility of corporate memory systems are presented.

FQL is a database query language in the area of functional programming systems and has been designed and developed in an attempt to provide a powerful formalism for the expression of complex database queries. There is no notion of data currency; full computational power is provided; complex queries may be developed incrementally; and the language itself is independent of any database management system. This document describes the syntactic and semantic rudiments of FQL through a set of examples. (Author)

The algorithm for synthesizing relational data base schema in the 3rd normal form assumes uniqueness of functional dependencies. This assumption is examined and a method for checking the assumption is presented. (Author)

The purpose of this document is to give a brief introduction to the hardware available in the Wharton Computer Center and the Decision Aiding Systems Laboratory (DASL), and to summarize the software which has been developed at Wharton under the Operational Decision Aiding Project. More detail about most of these features is available; in particular, some of the other software developed for use on the Wharton system may be of interest to other contractors, now that the system is easily accessible via the ARPANET. (Author)

This paper presents an architecture for a generalized model management system that facilitates the integration of management science models into a decision support system. The objective of the system is to support the decision-maker both in specifying a problem and in effecting a solution. This is accomplished by providing him/her with a means for interacting with a complex structured database to specify the structure of some problem; and to solve the model defined for the problem using appropriate information -- either from the database or some other source -- and efficient solution procedures.

This paper explores the concept of a database machine using the approach of the CODASYL data model. A database machine is defined as an integration of hardware and software for providing generalized database management capability in a physically separate device. The advantages of a database machine, along with functional specifications, are presented. Next, an illustration of using a database machine is given through an example of invoice processing using the SEED database management system on a DECsystem-10 computer. Finally, the implications of several design alternatives arising from the illustration are discussed. (Author)

This paper describes DBALERT, which performs simple alerting for the WAND DBMS. A demonstration is shown using DBALERT to perform alerting on the ONRODA and SYSFIL databases. (Author)

It is often desirable for programs to communicate by queueing messages in shared areas of secondary storage. An algorithm is presented for the allocation of variable size blocks which are to be freed in the same order that they are allocated. Unlike a wrap-around technique, it does not require an initial allocation of a fixed amount of memory, but exploits the ability of many operating systems to increase the available storage as needed. By means of a worst-case analysis, its space efficiency is shown to be optimal. (Author)

This paper describes the control programs and data structures used to implement multiple alerters in a relational data model environment. Alerters provide data base management systems with the capabilities of dynamically monitoring for the presence of user defined states of the data, and executing some predetermined action as a consequence of their detection. One of the main considerations in the implementation of an alerter system on a relational data base is the efficiency with which a change to a virtual relation can be detected. Programs are described for the definition, maintenance and evaluation of virtual relations. This work includes details of the use of construction diagrams as a means for their representation, as well as descriptions of the techniques for avoiding unnecessary recomputation and, when required, efficiently performing partial evaluation. (Author)

Because of the event-driven nature of asynchronous, concurrent processes, production systems (PS's) are an attractive modelling tool. The system of interest can be modelled with a large number of independent states, with independent actions, and the knowledge base can be conveniently encoded declaratively. However, asynchronous, concurrent processes normally have strict requirements for inter-process communication and co-ordination; this requires a substantial degree of inter-rule communication in the PS. The result of this is that a complex control structure is embedded in the short term memory(STM); this is generally considered unattractive for a number of reasons. This paper proposes a separate, explicit control structure for modelling asynchronous, concurrent processes with PS's. Specifically, the use of a Petri net is addressed.

An applicative language based upon recent ideas by John Backus has been developed. The language provides a powerful formalism for the expression of complex database queries. Though currently implemented with an interface to a CODASYL system, the language employs a sufficiently general data model that use with other database management systems is possible. This paper describes the language through a number of examples and outlines its implementation. (Author)

Traditional database systems only allow queries to be made about the data within the database, rather than about the data as well as the structure of the data. A set of primitive operations is proposed which allow a flexible means of entering and exploring the data and the structure of a database. (Author)

The Database Interface (DBI) is responsible for monitoring conditions on a data base managed by a particular type of Database Management System (DBMS). At the Wharton School there are several large databases which are managed by WAND, a CODASYL-like DBMS. This thesis outlines the design and implementation of a DBI for WAND. The DBI will be integrated into the Network Alerter Service (NAS) which is being designed and implemented as a general user service for ARPANET users. The NAS will allow the monitoring of databases at various sites on the ARPANET, for conditions of interest to the user. The WAND DBI is capable of efficiently monitoring these conditions or alerters, and resopnding appropriately when the previously specified condition occur. This monitoring must be done efficiently since the performance of the DBMS may otherwise become extremely degraded. (Author)

The advent of computer networking has made increasingly important the task of designing languages with which another program may talk to a database system. The DATACOMPUTER supports a query language which was designed to be generated by other programs. It maintains a quasi-relational database system with no direct linking between records. Q is an attempt to do the same thing for a network database: specifically SEED, which is a codasyl like system developed at the Wharton School. (Author)

Certain database query systems, especially those that interpret natural language, do not require the user to have comprehensive knowledge of the database schema in order to form a meaningful query. Instead, the query is used to reference entities in the database schema and a path is automatically found through the schema which connects those entities. In this paper, we will classify those schemata in which all paths are equivalent, and then derive a method of marking a schema in order to determine the natural paths between two entities when different paths give rise to different results. (Author)

This paper discusses the subject of 'alerters', which can be used in a database management system to provide the capability of informing a user when a specified state of the world (as reflected in the database) is reached. This paper describes recent research in building database management systems which incorporate alerting features, and provides a framework within which this and similar research efforts can be evaluated., Supersedes Rept. no. 75-12-04, AD-A037 437.

This thesis presents the DATA (Dynamic Alerting Transaction Analysis) System as an alternative to a conventional database management system. The DATA System contains no records corresponding to entities but rather is simply a time ordered list of transactions. The advantages of DATA in the areas of security, integrity, and operational ease is discussed and the concept of alerting is presented. An alerting system provides facilities to monitor changes to the database in order to perform some action whenever certain conditions become true. An alerter can be thought of as a program that continuously monitors the database and takes some specified action when the corresponding condition becomes true. The DATA system is an excellent tool to implement alerting due to its ability to view the database at previous points in time. The work describes the implementation of the DATA System.

This work is a partial implementation of a dynamic restructuring processor. The total dynamic restructuring processor allows several generations of database structure to coexist. Restructuring occurs incrementally as data is accessed in the database. Only restructuring of the conceptual schema (as defined by the ANSI/SPARC report) is considered on a CODASYL type database system called WAND. The implementation uses generation data structures which allow several related schemas and databases to coexist with proper restructuring translation done on the fly. The scope of this implementation includes an analysis of data requirements and general implementation strategy for the total processor and detailed design and programming of routines that provide run-time translation from one schema definition and associated database to a user using another but related schema definition. (Author)

DBLISP consists of a set of functions which provide a direct interface between the SEED network database system (version B04) and UCI LISP. Described here are instructions for using DBLISP, a sample session and various technical comments. (Author)

Relationships between conceptual schema design and internal schema design are considered. The author's earlier work on automatic derivation of conceptual schemata is reviewed, and a constrained optimization model for design of internal schemata is introduced. Trade-offs represented in the model are discussed, and a method for obtaining access frequency estimates is described. (Author)