Refine
Departments, institutes and facilities
Document Type
- Report (3)
- Conference Object (2)
- Article (1)
- Book (monograph, edited volume) (1)
- Doctoral Thesis (1)
Has Fulltext
- no (8)
Keywords
- Algorithmen (1)
- Algorithmus (1)
- Berechenbarkeit (1)
- Informationskomplexität (1)
- Komplexitätsklassen (1)
- Laufzeit-Komplexität (1)
- strukturelle Komplexität von Daten (1)
This report describes the design, the implementation and the usage of a system for managing different systems for automated theorem proving and automatically generated proofs. In particular, we focus on a user-friendly web-based interface and a structure for collecting and cataloguing proofs in a uniform way. The second point hopefully helps to understand the structure of automatically generated proofs and builds a starting point for new insights for strategies for proof planning.
User Modeling and Machine Learning for User Modeling have both become important research topics and key techniques in recent adaptive systems. One of the most intriguing problems in the `information age´ is how to filter relevant information from the huge amount of available data. This problem is tackled by using models of the user´s interest in order to increase precision and discriminate interesting information from un-interesting data. However, any user modeling approach suffers from several major drawbacks: User models built by the system need to be inspectable and understandable by the user himself. Secondly, users in general are not willing to give feedback concerning user satisfaction by the delivered results.
As soon as data is noisy, knowledge as it is represented in an information system becomes unreliable. Features in databases induce equivalence relations—but knowledge discovery takes the other way round: given a relation, what could be a suitable functional description? But the relations we work on are noisy again. If we expect to record data for learning a classification of objects then it can well be the real data does not create a reflexive, symmetric and transitive relation although we know it should be. The usual approach taken here is to build the closure in order to ensure desired properties. This, however, leads to overgeneralisation rather quickly.
Formal concept analysis (FCA) as introduced in [4] deals with contexts and concepts. Roughly speaking, a context is an environment that is equipped with some kind of "knowledge". Such contexts are also known as information or knowledge representation systems where the knowledge consists of (intensional) descriptions relating sets of objects to sets of properties. Given extsensional and intensional descriptions (the latter one in terms of binary attributes), they can be arranged in a taxonomy or concept lattice.
This report summarises and integrates two different tracks of research for the purpose of envisioning and preparing a joint research project proposal. Soft- and hardware systems have become increasingly complex and act "concurrently", both with respect to memory access (i.e. information flow) and computational resources (i.e. "services"). The software development metaphor of cloud-storage, cloud-computing and service-oriented design has been anticipated by artificial intelligence (AI) research at least 30 years ago (parallel and distributed computation already dates back to the 1950’s and 1970s). What is known as a "service" today is what in AI is known as the capability of an agent; and the problem of information flow and consistency has been a headstone of information processing ever since. Based on a real-world robotics application we demonstrate how an increasingly abstract description of collaborating or competing agents correspond to a set of concurrent processes.