### Refine

#### Year of publication

- 1999 (397) (remove)

#### Document Type

- Preprint (397) (remove)

#### Keywords

- Case-Based Reasoning (10)
- Fallbasiertes Schliessen (5)
- Location Theory (5)
- case-based problem solving (5)
- Abstraction (4)
- Fallbasiertes Schließen (4)
- Knowledge Acquisition (4)
- Internet (3)
- Knowledge acquisition (3)
- Maschinelles Lernen (3)

#### Faculty / Organisational entity

Let rC and rD be two convexdistance funtions in the plane with convex unit balls C and D. Given two points, p and q, we investigate the bisector, B(p,q), of p and q, where distance from p is measured by rC and distance from q by rD. We provide the following results. B(p,q) may consist of many connected components whose precise number can be derived from the intersection of the unit balls, C nd D. The bisector can contain bounded or unbounded 2-dimensional areas. Even more surprising, pieces of the bisector may appear inside the region of all points closer to p than to q. If C and D are convex polygons over m and m vertices, respectively, the bisector B(p,q) can consist of at most min(m,n) connected components which contain at most 2(m+n) vertices altogether. The former bound is tight, the latter is tight up to an additive constant. We also present an optimal O(m+n) time algorithm for computing the bisector.

In planar location problems with barriers one considers regions which are forbidden for the siting of new facilities as well as for trespassing. These problems areimportant since they reflect various real-world situations.The resulting mathematical models have a non-convex objectivefunction and are therefore difficult to tackle using standardmethods of location theory even in the case of simple barriershapes and distance funtions.For the case of center objectives with barrier distancesobtained from the rectilinear or Manhattan metric it is shown that the problem can be solved by identifying a finitedominating set (FDS) the cardinality of which is bounded bya polynomial in the size of the problem input. The resultinggenuinely polynomial algorithm can be combined with bound computations which are derived from solving closely connectedrestricted location and network location problems.It is shown that the results can be extended to barrier center problems with respect to arbitrary block norms having fourfundamental directions.

In this paper we deal with an NP-hard combinatorial optimization problem, the k-cardinality tree problem in node weighted graphs. This problem has several applications , which justify the need for efficient methods to obtain good solutions. We review existing literature on the problem. Then we prove that under the condition that the graph contains exactly one trough, the problem can be solved in ploynomial time. For the general NP-hard problem we implemented several local search methods to obtain heuristics solutions, which are qualitatively better than solutions found by constructive heuristics and which require significantly less time than needed to obtain optimal solutions. We used the well known concepts of genetic algorithms and tabu search with useful extensions. We show that all the methods find optimal solutions for the class of graphs containing exactly one trough. The general performance of our methods as compared to other heuristics is illustrated by numerical results.

Given a finite set of points in the plane and a forbidden region R, we want to find a point X not an element of int(R), such that the weighted sum to all given points is minimized. This location problem is a variant of the well-known Weber Problem, where we measure the distance by polyhedral gauges and allow each of the weights to be positive or negative. The unit ball of a polyhedral gauge may be any convex polyhedron containing the origin. This large class of distance functions allows very general (practical) settings - such as asymmetry - to be modeled. Each given point is allowed to have its own gauge and the forbidden region R enables us to include negative information in the model. Additionally the use of negative and positive weights allows to include the level of attraction or dislikeness of a new facility. Polynomial algorithms and structural properties for this global optimization problem (d.c. objective function and a non-convex feasible set) based on combinatorial and geometrical methods are presented.

In this paper a new trend is introduced into the field of multicriteria location problems. We combine the robustness approach using the minmax regret criterion together with Pareto-optimality. We consider the multicriteria Weber location problem which consists of simultaneously minimizing a number of weighted sum-distance functions and the set of Pareto-optimal locations as its solution concept. For this problem, we characterize the Pareto-optimal solutions within the set of robust locations for the original weighted sum-distance functions. These locations have both the properties of stability and non-domination which are required in robust and multicriteria programming.

The problem of finding an optimal location X* minimizing the maximum Euclidean distance to existing facilities is well solved by e.g. the Elzinga-Hearn algorithm. In practical situations X* will however often not be feasible. We therefore suggest in this note a polynomial algorithm which will find an optimal location X^F in a feasible subset F of the plane R^2

In the following, we discuss a procedure for interpolating a spatial-temporal stochastic process. We stick to a particular, moderately general model but the approach can be easily transered to other similar problems. The original data, which motivated this work, are measurements of gas concentrations (SO2, NO, O2) and several meteorological parameters (temperature, sun radiation, procipitation, wind speed etc.). These date have been and are still recorded twice every hour at several irregularly located places in the forests of the state Rheinland-Pfalz as part of a program monitoring the air pollution in the forests.

In this paper we deal with the location of hyperplanes in n-dimensional normed spaces. If d is a distance measure, our objective is to find a hyperplane H which minimizes f(H) = sum_{m=1}^{M} w_{m}d(x_m,H), where w_m ge 0 are non-negative weights, x_m in R^n, m=1, ... ,M demand points and d(x_m,H)=min_{z in H} d(x_m,z) is the distance from x_m to the hyperplane H. In robust statistics and operations research such an optimal hyperplane is called a median hyperplane. We show that for all distance measures d derived from norms, one of the hyperplanes minimizing f(H) is the affine hull of n of the demand points and, moreover, that each median hyperplane is (ina certain sense) a halving one with respect to the given point set.

There are several good reasons to introduce classification schemes for optimization problems including, for instance, the ability for concise problem statement opposed to verbal, often ambiguous, descriptions or simple data encoding and information retrieval in bibliographical information systems or software libraries. In some branches like scheduling and queuing theory classification is therefore a widely accepted and appreciated tool. The aim of this paper is to propose a 5-position classification which can be used to cover all location problems. We will provide a list of currentliy available symbols and indicate its usefulness in a - necessarily non-comprehensive - list of classical location problems. The classification scheme is in use since 1992 and has since proved to be useful in research, software development, classroom, and for overview articles.

We present an approach to learning cooperative behavior of agents. Our ap-proach is based on classifying situations with the help of the nearest-neighborrule. In this context, learning amounts to evolving a set of good prototypical sit-uations. With each prototypical situation an action is associated that should beexecuted in that situation. A set of prototypical situation/action pairs togetherwith the nearest-neighbor rule represent the behavior of an agent.We demonstrate the utility of our approach in the light of variants of thewell-known pursuit game. To this end, we present a classification of variantsof the pursuit game, and we report on the results of our approach obtained forvariants regarding several aspects of the classification. A first implementationof our approach that utilizes a genetic algorithm to conduct the search for a setof suitable prototypical situation/action pairs was able to handle many differentvariants.

The common wisdom that goal orderings can be used to improve planning performance is nearly as old as planning itself. During the last decades of research several approaches emerged that computed goal orderings for different planning paradigms, mostly in the area of state-space planning. For partial-order, plan-space planners goal orderings have not been investigated in much detail. Mechanisms developed for statespace planning are not directly applicable because partial-order planners do not have a current (world) state. Further, it is not completely clear how plan-space planners should make use of goal orderings. This paper describes an approach to extract goal orderings to be used by the plan-space planner CAPlan. The extraction of goal orderings is based on the analysis of an extended version of operator graphs which previously have been found useful for the analysis of interactions and recursion of plan-space planners.

Im Rahmen des Sonderforschungsbereichs SFB314, Projekt X9 "Lernen und Analogie in technischen Expertensystemen", wurde die Verwendbarkeit von Techniken des fallbasierten Schliessens in wissens- basierten Systemen untersucht. Als prototypische Anwendungsdomäne wurde die Arbeitsplanerstellung rotationssymmetrischer Werkstücke gewählt. Im vorliegenden Beitrag wird ein Modell der Arbeits- planerstellung unter Berücksichtigung der verschiedenen, bisher als unabhängig behandelten Planungsmethoden beschrieben. Auf der Basis einer modelbasierte Wissensakquistion aus in Unternehmen verfügbaren Arbeitsplänen wird ein Ausschnitt der Arbeitsplanerstellung, die Aufspannplanung, detailliert. Die Anwendbarkeit wurde durch eine prototypische Realisierung nachgewiesen.

Freivalds, Karpinski and Smith [8] explored a special type of learning in the limit: identification of an unknown concept (function) by eliminating (erasing) all but one possible hypothesis (this type of learning is called co-learning). The motivation behind learning by erasing lies in the process of human and automated computer learning: often we can discard incorrect solutions much easier than to come up with the correct one. In Gödel numberings any learnable family can be learned by an erasing strategy. In this paper we concentrate on co-learning minimal programs. We show that co-learning of minimal programs, as originally defined is significantly weaker than learning minimal programs in Gödel numberings. In order to enhance the learning power

We present an approach to automating the selection of search-guiding heuris-tics that control the search conducted by a problem solver. The approach centerson representing problems with feature vectors that are vectors of numerical val-ues. Thus, similarity between problems can be determined by using a distancemeasure on feature vectors. Given a database of problems, each problem beingassociated with the heuristic that was used to solve it, heuristics to be employedto solve a novel problem are suggested in correspondence with the similaritybetween the novel problem and problems of the database.Our approach is strongly connected with instance-based learning and nearest-neighbor classification and therefore possesses incremental learning capabilities.In experimental studies it has proven to be a viable tool for achieving the finaland crucial missing piece of automation of problem solving - namely selecting anappropriate search-guiding heuristic - in a flexible way.This work was supported by the Deutsche Forschungsgemeinschaft (DFG).

This report presents the properties of a specification of the domain of process planning for rotary symmetrical workpieces. The specification results from a model for problem solving in this domain that involves different reasoners, one of which is an AI planner that achieves goals corresponding to machining workpieces by considering certain operational restrictions of the domain. When planning with SNLP (McAllester and Rosenblitt, 1991), we will show that the resulting plans have the property of minimizing the use of certain key operations. Further, we will show that, for elastic protected plans (Kambhampati et al., 1996) such as the ones produced by SNLP, the goals corresponding to machining parts of a workpiece are OE-constrained trivial serializable, a special form of trivial serializability (Barrett and Weld, 1994). However, we will show that planning with SNLP in this domain can be very difficult: elastic protected plans for machining parts of a workpiece are nonmergeable. Finally, we will show that, for sufix, prefix or sufix and prefix plans such as the ones produced by state-space planners, it is not possible to have both properties, being OEconstrained trivial serializable and minimizing the use of the key operations, at the same time.

In nebenläufigen Systemen erleichtert das Konzept der Atomarität vonOperationen, konkurrierende Zugriffe in größere, leichter beherrschbareAbschnitte zu unterteilen. Wenn wir aber Spezifikationen in der forma-len Beschreibungstechnik Estelle betrachten, erweist es sich, daß es un-ter bestimmten Umständen schwierig ist, die Atomarität der sogenanntenTransitionen bei Implementationen exakt einzuhalten, obwohl diese Ato-marität eine konzeptuelle Grundlage der Semantik von Estelle ist. Es wirdaufgezeigt, wie trotzdem sowohl korrekte als auch effiziente nebenläufigeImplementationen erreicht werden können. Schließlich wird darauf hinge-wiesen, daß die das Problem auslösenden Aktionen oft vom Spezifiziererleicht von vorneherein vermieden werden können; und dies gilt auch überden Kontext von Estelle hinaus.

Top-down and bottom-up theorem proving approaches have each specific ad-vantages and disadvantages. Bottom-up provers profit from strong redundancycontrol and suffer from the lack of goal-orientation, whereas top-down provers aregoal-oriented but have weak calculi when their proof lengths are considered. Inorder to integrate both approaches our method is to achieve cooperation betweena top-down and a bottom-up prover: The top-down prover generates subgoalclauses, then they are processed by a bottom-up prover. We discuss theoreticaspects of this methodology and we introduce techniques for a relevancy-basedfiltering of generated subgoal clauses. Experiments with a model eliminationand a superposition-based prover reveal the high potential of our cooperation approach.The author was supported by the Deutsche Forschungsgemeinschaft (DFG).