Informatik-Kolloquium WiSe 2009/2010

Termine, Vortragende, Titel und Zusammenfassungen für das Kolloquium im Wintersemester 2009/2010.

Die Vorträge finden in unregelmäßigen Abständen in der Regel Freitags um 14:00 Uhr im Seminarraum 049 in der Takustr. 9 statt.

Vorläufige Terminreservierungen ohne Gastname sind als "(reserviert von [Einladende/r])" einzutragen; Termine noch ohne Gewähr als "(wahrscheinlich)" o.ä. zu markieren. Im Rumpf des Eintrags steht jeweils, wer den Vortrag organisiert. (Zur technischen Notation siehe ShortHand.)

Ab spätestens einen Tag vor dem Vortrag sollte auch eine Zusammenfassung dabeistehen.

Information zum Anlegen von Kolloquiumseiten pro Semester findet man unter KolloWeitereInfos

September

Fr, 18.09.2009: Dialogue Model Specification and Interpretation for Interacting with Service Robots, Luis Pineda, Jefe del Departamento de Ciencias de la Computación IIMAS Nationale Universität Mexikos, (Einladender: R. Rojas)

Takustr. 9, Raum 006, 14:15 bis ca. 15:45

Abstract: In this talk a notion of multimodal dialogue models specification and interpretation for interacting with service robots is presented. First, some general considerations regarding the design of situated agents with coordinated language, vision and motor capabilities are discussed. It is emphasized the need to adopt a wide notion of interaction context, and to model it in a simple and general way. Then an architecture for intelligent situated agents is presented. It consists of three main functional layers that we name recognition, interpretation and representation. In this architecture, perception is guided by the system's linguistic and visual expectations. Then the notion of dialogue model is introduced. Dialogue models are representations of the interaction context for specific application domains. A formalism for representation and interpretation of these objects is also presented. Dialogue models are expressed in a declarative way and their interpretation proceeds in tandem with the interaction between the agent and the world. The present theory and methodology has been used for the design of a conversational robot, which is also presented. The talk is concluded with the illustration of some applications constructed in this framework.

---

Oktober

Fr, 30. 10. 2009: (Geplant) Prof. Dr. Tetsuo Asano, Japan Advanced Institute of Science and technology (JAIST), (Einladender: C. Knauer)

TITLE: Constant-Working Space Algorithms: How Fast can we Solve Problems Without Using Extra Array

ABSTRACT: In this talk I will talk about a new direction of algorithms which do not use any extra working array. More formally, we want to design efficient algorithms which require no extra array of size depending on input size n but use only constant working storage cells (variables), each having O(log n) bits. This talk covers some problems related to geometry and also binary images. Here is a simple problem. Given a tree and two arbitrary nodes s and t, report a simple path from s to t in order. This problem is easy if a working array is available. Can you design an algorithm without using any extra array and assuming $T$ is given on a read-only array?

Takustr. 9, Raum 049, 14:15 bis ca. 15:45

---

November

Fr, 20. 11. 2009: Andy King (University of Kent, England), (Einladender: G. Rote)

TITLE: Logahedra – a Geometric Shape for Bug Finding

ABSTRACT: The field of program analysis is concerned with extracting information from a program, without human intervention, in order to support compiler optimization or code transformation. These days, when there is less emphasis on speed and more on correctness, program analyses are being increasingly reinterpreted as methods for automating bug discovery. For instance, if an array is accessed out of range, then the program is faulty. Thus there is interest in applying analyses to infer the range of values that variables can take before the program is executed, to provide a guarantee of (partial) correctness. This talk will show how geometric objects called logahedra address some of the representational problems that arise in analysis. The work will be illustrated by running example that visually illustrates how geometry promises to be important in bug finding.

Takustr. 9, Raum 049, 14:15 bis ca. 15:15 -- Main.rote - 29 Sep 2009

---

Fr, 27. 11. 2009: Disputation Astrid Sturm im Zib-Seminarram (Einladender: G. Rote). daher diesen Termin lieber freilassen.-- Main.rote - 11 Nov 2009

Fr, 27. 11. 2009: Gerald Weber, University of Auckland, (Einladender: R. Rojas) [Vortragender kann nur an diesem Termin, deshalb trotz des Disputationsvortrags]

TITLE: Dokumentenorientierte Benuzterschnittstellen

ABSTRACT: Neuere Bibliotheken für Benutzerschnittstellen verwenden zunehmend Beschreibungssprachen wie XUL und XAML. Diese Veränderung ist zur Zeit aber nur eine technische Restrukturierung und bietet nur begrenzten Mehrwert für den Endbenutzer. In unserer Forschungsgruppe haben wir seit einigen Jahren eine konzeptionelle Theorie von Dokumentenorientierung entwickelt und diese in vielfältigen Projekten implementiert, die auf die Vision des Auckland Interface Models (AIM) hinlaufen.

[1] LUTTEROTH, C., STRANDH, R., WEBER, G. 'Domain Specific High-Level Constraints for User Interface Layout', Constraints Journal (2008), 13:307-342 Springer. [2] CHAKRAVARTHI, Y. A., LUTTEROTH, C., WEBER, G. 'AIMHelp: Generating Help for GUI Applications Automatically', Proceedings of the 10th international Conference NZ Chapter of the Acm's Special interest Group on Human-Computer interaction (Auckland, New Zealand, July 06 - 07, 2009). CHINZ '09. ACM, New York, NY, 21-28.

Die Vortragsunterlagen sind in englischer Sprache verfasst.

---

Dezember

---

Januar

Mo, 25.1.2010: (GEPLANT) Prof. R. Kruse, Univ.Magdeburg, Einladender: H. Schweppe

Takustr. 9, Raum 049, 16:15 bis ca. 16:45

---

Februar

Fr, 12.2.2010: Prof. H. Hermanns, Universität des Saarlandes, Einladender: M. Kyas

Takustr. 9, Raum 049, 14:15 bis ca. 15:45

TITLE: Models & Analysis Methods for Sensors Networks

ABSTRACT: The design of wireless sensor network (WSN) is fun. But reliable designs are difficult to arrive at for at least three reasons: (1) inherent concurrency, (2) clock inaccuracies and drift, and (3) interaction with and through noisy environments. It is therefore virtually impossible to achieve good lab conditions even inside a lab: experimental observations are notoriously difficult to reproduce. Therefore, the only way forward is to study mathematical models of sensor network designs, to distill abstract but generally valid properties of a design.

After sketching the problem domain, this talk reviews the most relevant model classes in this context, starting off from automata models as they appear in the verification of real-time and hybrid systems, and from stochastic processes used in classical computer performance evaluation. We then lay out the principal ingredients of the model of stochastic (priced) timed automata, which has suitable theoretical and pragmatic properties. A modelling language and its compositional semantics are presented, and applied to some exemplary aspects from the WSN domain. We finish with a discussion of the status and the prospect of tool support.

---

Vorlage

Fr, xx.xx.2008: Titel

Vortragende/r, Herkunftsorganisation, (eingeladen von Einladende/r)

Takustr. 9, Raum xx, 14:15 bis ca. 15:45

Zusammenfassung

Kolloquium in anderen Semestern

Nächstes Semester

InformatikKolloquiumSoSe2010

Frühere Semester

InformatikKolloquiumSoSe2009
InformatikKolloquiumSoSe2008
InformatikKolloquiumWiSe2008
InformatikKolloquiumSoSe2007
InformatikKolloquiumWiSe2006
InformatikKolloquiumSoSe2006
InformatikKolloquiumWiSe2005
InformatikKolloquiumSoSe2005

---

Kommentare