17 June 2011
STP'11 Workshop, University of Dundee
The web-page of the workshop is here.
03 November 2010
Simon Dobson and Juan Ye
Title: Sensor and sense-abilty: building systems in the face of uncertainty
Abstract: It has been an old maxim in computing that incorrect inputs can acceptably give rise to unacceptable outputs: "garbage in, garbage out". This is ceasing to be true, and many classes of systems must behave predictably even in the face of inputs containing substantial garbage -- although researchers delicately use terms like "imprecise" or "unstructured" instead of "garbage". In this talk we discuss some approaches to managing the problem of imprecise, inaccurate, untimely and partial inputs in the context of pervasive and sensor-driven systems, and suggest that we need to re-think radically the way we build software and represent decision-making in these environments.
6 October 2010:
Dr Ewen Maclean, Heriot-Watt University.
Invited Lecture on Software Verification. School of Computing, University of Dundee.
Why computers need to prove themselves.
Abstract: Recent news stories such as the serious faults in the Airbus flight software, and the apparent failing of the infamous Toyota accelerator pedal show that correctness in computer programs is a critical safety concern. As opposed to rigorous testing, it is possible to use Formal Methods to specify and prove properties about computer programs use Automated Theorem Provers, thus guaranteeing their safety. In this talk I will explain some of the original motivations behind software verification, and discuss some modern and successful techniques which have reignited interested and fruitful research in the area. In particular I will focus on the ways in which inventive Artifiical Intelligence techniques can be employed to recreate "Eureka" steps in mathematical proofs which have in the past rendered formal proof of programs infeasible. Some aspects of the talk will be technical and involve mathematics, but I will endeavour to keep the motivations clear.
July-August 2010:
CLANN hosted SICSA visiting fellow Dr John Power.
Series of lectures at the School of Computer Science, University of Strathclyde, and a lecture at the School of Computing, university of Dundee.
10 August 2010.
Title: An Historical Introduction to Category Theory and one of its roles in Theoretical Computer Science for those who have never heard of it
Abstract: In this talk, or preferably discussion, I shall explain some of the historical setting in which category theory developed, and trace an historical strand from there through to one of its more prominent uses in theoretical computer science. This amounts to one of the lines of development of a topic within pure mathematics called universal algebra. The aim of the talk is to present human, historical development rather than technical work; and I would welcome contributions by anyone present who might have a different perspective on the history to that I have found.
15 July 2010
PAR'10 Workshop at FLOC'10, Edinburgh
The page of the workshop is here.
15-20 March 2009:
Prof. Dr Kai-Uwe Kuehnberger, Institute of Cognitive Science, University of Osnabruck.
Kai-Uwe Kuehnberger is an assistant professor for artificial intelligence at the Institute of Cognitive Science (IKW) of the University of Osnabrueck (Germany). He obtained a Master's degree from the University of Tuebingen with a thesis about fixed points in algebraic structures and a PhD in Computational Linguistics with a thesis about modeling circularity in natural and formal languages. His main research interest is the modeling of higher cognitive abilities. In particular, he is interested in analogical reasoning, ontology design and text technology, neural-symbolic integration, and cognitive architectures. He is one of the executive editors of the newly established Journal of Artificial General Intelligence, survey editor of the Elsevier Journal Cognitive Systems Research, and series editor of the newly established book series "Thinking Machines" (Atlantis Press). He has been serving as co-PI in the project "Modeling of Analogies with Heuristic-Driven Theory Projection" and in the project "Adaptive Ontologies on Extreme Markup Languages", both funded by the German Research Foundations. Furthermore, he is one of the co-PIs of the graduate school "Cognitive Science" at the IKW funded by the state Lower Saxony.
Kai-Uwe will give us two talks, as follows:
17 March 2009
Kai-Uwe Kuehnberger: Natural Computation for Reasoning and Learning Applications
Abstract:
Mainstream reasoning and learning techniques have several deficiencies. They are rather cognitively implausible, there are too many different types of reasoning, learning, and corresponding representation formalisms, and learning from sparse and noisy data often challenge classical approaches in practical applications, just to mention some of these problems. A way out of these problems could be the usage of new computing paradigms. This presentation will give some ideas (clearly no ready-made solutions) for cognitively motivated and biologically inspired candidates of natural computation to resolve some of the mentioned problems: DOM tree learning with kernel methods can be used to classify text genres, neural-symbolic integration can be a tool to learn logical inferences with neural networks, and analogical reasoning is a candidate for the integration of various reasoning types for computing the meaning of metaphoric expressions. Promising application domains for such approaches may be text technology, semantic web, and ontology design.
18 March 2009
Kai-Uwe Kuehnberger: Translating Logic into Neural Networks
Abstract:
There is an obvious tension between systems for symbolic processing and systems for subsymbolic processing. Both have complementary strengths and weaknesses in applications and the underlying methodologies. In order to resolve this theoretically unsatisfactory and practically undesirable situation we propose the translation of logic theories into a Topos representation, i.e. we propose a translation of logic into commuting diagrams (Goldblatt, 1979). The Topos representation codes not only the axioms of a given logical theory, but also possible inferences of such an axiomatic system due to Topos constructions like products or pullbacks. Such constructions can be used for training neural networks. The network learns a particular model of the axiomatic system, more precisely, it learns representations of constants, predicates, truth values of formulas etc. I will present some of the features of this system and discuss some of the occurring problems.
Resources:
20-23 January 2009:
Dr Pascal Hitzler, Institute for Applied Computer Science and Formal Description Methods, University of Karlsruhe.
PD Dr. Pascal Hitzler is assistant professor at the Institute for Applied Informatics and Formal Description Methods (AIFB) at the University of Karlsruhe in Germany since 2004. Before this, he worked as a postdoc at the Artificial Intelligence Institute at TU Dresden. His research record lists over 120 publications in such diverse areas as semantic web, neural-symbolic integration, knowledge representation and reasoning, lattice and domain theory, denotational semantics, and set-theoretic topology. At AIFB, he is supervising research on knowledge representation and reasoning in several large-scale projects funded by the German Federal Ministry for Education and Research and the European Commission. He is on the editorial board of the Journal on Advances in Artificial Intelligence, of the Journal of Algorithms in Cognition, Informatics and Logic, and of the Journal of Artificial General Intelligence. He is in the steering committees of the International Conference on Conceptual Structures, of the conference on Web Reasoning and Rule Systems (as vice-chair), and of the workshop series OWL - Experiences and Directions. He also serves as a member of the W3C OWL working group. He is co-author of the first german introductory texbook to the semantic web published by Springer Verlag, and he has co-edited a Springer book on Pespectives of Neural-Symbolic Integration. He has invented and been co-chair of the workshop series on Neural-symbolic Learning and Reasoning (NeSy) which has so far been held at IJCAI-05, ECAI-06, IJCAI-07, and ECAI-08, and will be held at IJCAI-09. For more information, see http://www.pascal-hitzler.de.
Pascal gave us two talks, as follows:
21 January 2009
Pascal Hitzler: Knowledge Representation and Reasoning for the Semantic Web
Abstract: The emerging Semantic Web is based on the enrichment of data on the World Wide Web by so-called metadata in the form of ontologies, which are used for conveying the meaning, or semantics, of data on the Web. The techniques enable the use of intelligent systems on the Web for applications such as enhanced search, knowledge integration, ubiquitous computing, and many others. Foundational for the Semantic Web is the representation of knowledge by logic-based formalisms, together with suitable automated reasoning techniques. For this purpose, the World Wide Web Consortium has established recommendations for representation standards, including the Web Ontology Language OWL which is based on Description Logics. In this talk, we will present major knowledge representation formalisms for the Semantic Web and discuss recent challenges and developments. In particular, we will present approaches to dealing with the problem of scalability of automated reasoning and new results on the integration of ontologies and rule-based systems.
References:
Markus Krotzsch, Sebastian Rudolph, Pascal Hitzler, ELP: Tractable Rules for OWL 2. In: A Sheth, Steffen Staab, Mike Dean, Massimo Paolucci, Diana Maynard, Timothy Finin, Krishnaprasad Thirunarayan (eds.), The Semantic Web - ISWC 2008, 7th International Semantic Web Conference. Springer Lecture Notes in Computer Science Vol. 5318, 2008, pp. 649-664.
Markus Krotzsch, Sebastian Rudolph, Pascal Hitzler, Description Logic Rules. In: Malik Ghallab, ConstantineD. Spyropoulos, Nikos Fakotakis, Nikos Avouris (eds.), Proceedings of the 18th European Conference on Artificial Intelligence, ECAI2008, Patras, Greece, July 2008. IOS Press, 2008, pp. 80-84.
Tuvshintur Tserendorj, Sebastian Rudolph, Markus Krotzsch, Pascal Hitzler, Approximate OWL-Reasoning with Screech. In: Diego Calvanese, Georg Lausen (eds.), Web Reasoning and Rule Systems, Second International Conference, RR 2008, Karlsruhe, Germany, October/November 2008. Springer Lecture Notes in Computer Science Vol. 5341, 2008, pp. 165-180.
Pascal Hitzler, Markus Krotzsch, Sebastian Rudolph, Foundations of Semantic Web Technologies. Textbooks in Computing, Chapman and Hall/CRC Press, 2009.
22 January 2008
Pascal Hitzler: Neural-symbolic Integration
Abstract: Intelligent systems based on symbolic knowledge processing, on the one hand, and on artificial neural networks (also called connectionist systems), on the other, differ substantially. Nevertheless, both of these are standard paradigms in artificial intelligence, and it would be very desirable to combine the robustness of neural networks with the expressivity of symbolic knowledge representation. This is the reason why the importance of the efforts to bridge the gap between the connectionist and symbolic paradigms of Artificial Intelligence is widely recognised and gaining momentum. As the amount of hybrid data containing symbolic and statistical elements as well as noise increases in diverse areas such as bioinformatics or text and web mining, neural-symbolic learning and reasoning becomes of particular practical importance. In this talk we report on past and recent developments in this area, which are taylored towards the connectionist acquisition and processing of symbolic knowledge. Particular emphasis will be given to the problem of connectionist treatments of knowledge which goes beyond propositional logic, e.g. in the form of first-order predicate logic programs. This includes a report on joint work with Sebastian Bader, Steffen Hölldobler and Andreas Witzel on the design and implementation of an integrated neural-symbolic learning system which can process such logic programs.
References:
Sebastian Bader, Pascal Hitzler, Steffen Holldobler, Connectionist Model Generation: A First-Order Approach. Neurocomputing 71, 2008, 2420-2432.
Sebastian Bader, Pascal Hitzler, Steffen Holldobler, Andreas Witzel, A Fully Connectionist Model Generator for Covered First-Order Logic Programs. In: Manuela M. Veloso, Proceedings of the Twentieth International Joint Conference on Artificial Intelligence, IJCAI-07, Hyderabad, India, January 2007, AAAI Press, Menlo Park CA, 2007, pp. 666-671.
Barbara Hammer, Pascal Hitzler (eds.), Perspectives of Neural-Symbolic Integration. Studies in Computational Intelligence, Vol. 77. Springer, 2007.