The Eleventh International Workshop on Qualitative Reasoning

Article excerpt

The Eleventh International Workshop on Qualitative Reasoning was held in Cortona, Italy, on 3 to 6 June 1997. Participants included scientists from both qualitative reasoning and quantitative mathematical modeling communities. This article summarizes the significant issues and discussion raised during the workshop.

The Eleventh International Workshop on Qualitative Reasoning (QR '97) was held on 3 to 6 June 1997 at Il Palazzone, an imposing and regal sixteenth-century villa in the medieval town of Cortona, Tuscany, Italy. Sixty-four people participated, and 39 papers were presented in either oral or poster sessions.

The meeting was sponsored by Istituto di Analisi Numerica (IAN) - C.N.R., the American Association for Artificial Intelligence, IllyCaffe, the Italian Association for Artificial Intelligence, OCC'M Software GmbH, Stigma s.r.l Pavia, and Xerox Palo Alto Research Center. The support that was also given by academic and industrial institutions that are not traditionally devoted to AI research but, rather, to modeling methodologies and their applications highlights the intrinsic interdisciplinary nature of qualitative reasoning research. However, the evolution of qualitative reasoning research to increasingly sophisticated methods and techniques, along with their application to increasingly complex domains, inevitably requires a strong interaction with those communities that have traditionally been devoted to mathematical modeling.

QR '97 broke with tradition in that it was organized by an institution, the Istituto di Analisi Numerica - CNR of Pavia, whose main focus is quantitative mathematical modeling in the domain of applied sciences, although a research project on qualitative reasoning methods and their integration with quantitative reasoning methods has been active since 1988. Given the organizational context, an additional goal in our minds in preparing the workshop was to establish a basis for interaction between the qualitative and quantitative communities.

To this end, in addition to the presentation of full papers, posters, and short talks, in line with past workshop schedules, we planned invited talks, the focuses of which were problem domains for qualitative reasoning in real-world applications, and a tutorial on system identification. In particular, Furio Suggi-Liverani (IllyCaffe - Trieste) discussed the problems related to the formulation of a quantitative model of the espresso coffee brewing dynamics; G. Zanetti and G. Fotia (CRS4 - Cagliari) focused on problems of interpreting massive data streams coming from latest-generation medical-imaging equipment and of modeling of liquid-solid interaction, respectively. A tutorial entitled "System Identification: Problems and Perspectives" was presented by G. De Nicolao (University of Pavia). The tutorial gave an overview of the main problems (identifiability, overparameterization, model comparison) arising in system identification from data. The tutorial was appreciated by the participants because this topic is of fundamental importance in automated modeling.

A valuable impact of qualitative reasoning to system identification is confirmed by three papers presented at the workshop. E. Bradley, A. O' Gallagher, and J. Rogers (all of the University of Colorado) presented a tool that uses qualitative information to improve parameter-estimation techniques. This work is part of a more general tool for automated system identification and aims at finding optimal parameter choices to match a nonlinear ordinary differential equation model to data without getting trapped in local minima. S. Reece (University of Oxford) addressed the problem of data fusion and parameter estimation for noisy processes when the process models are incomplete or imprecise. The noisy data are filtered by a qualitative Kalman filter proposed by the author. The paper by A. Steele and R. Leitch (both of University of Edinburgh) addressed the problem of qualitative parameter identification and presented an exhaustive search technique to estimate parameters in a fuzzy qualitative model. …