ERCIM News No.45 - April 2001 [contents]
Working Group on Soft Computing proposed
by Petr Hájek
Soft Computing is an establishing domain and researchers from ERCIM insitutes propose to form an working group in this field.
Soft Computing has emerged as an attempt to formulate a new paradigm of computing. It can be discribed as association of computing methodologies centering on fuzzy logic, neurocomputing, genetic computing, and probabilistic computing. The guiding principle of soft computing is: exploit the tolerance for imprecision, uncertainty, partial truth, and approximation to achieve tractability, robustness, low solution cost and better rapport with reality. One of the principal aims of soft computing is to provide a foundation for the conception, design and application of intelligent systems employing its member methodologies symbiotically rather than in isolation. Scientists from ERCIM insitutes propose to form an working group in this field, comprising the following research topics:
Mathematical and logical foundations of Soft Computing
- systematic development of the theoretical basis of non-standard (ie approximate, nondeterministic or uncertain, etc.) reasoning, with a special focus to approaches to fuzzy logic, probabilistic and possibilistic logic, Dempster-Shafer theory and related approaches
- research in fuzzy logic programming in predicate calculus and with wider class of connectives (conjunctors and aggregation operators) which appear in practical application where one works with approximation of connectives
- design of the respective calculi for quantification and processing of uncertainty, imprecision and vagueness, analysis of their theoretical properties both from logical and algorithmic points of view; mutual confrontation of these calculi from the viewpoint of their expressive and processing power
- development of the theory of complexity of feed forward and recurrent neural networks and their learning algorithms based on complexity measure corresponding to various implementation possibilities.
Algorithmical Foundations of Soft Computing
- design and investigation of models suitable for realisation of formal calculi studied/proposed in the previous item
- design, development and analysis of formal, abstract machine models embodying the ideas of soft computing, inspired by biological or genetic models, with special regard to massively parallel models, distributed models, and neuromorphic models
- extending the theoretical basis of neural network-based computation with a special regard to approximation theory and related development of new neural computation paradigms
- the design and analysis of efficient algorithms for the fundamental problems in soft computing, both for internal need of the theory, as well as for various application areas, especially in the field of data mining.
Experimental Applications of Soft Computing
- identifying new soft computing information processing application areas particularly in the field of fuzzy and neuro-fuzzy systems, hybrid (ie, analog and neural) systems, data/knowledge bases, data warehouses, data mining, etc.
- identification and formalisation (modelling) of paradigmatical problems of Soft Computing
- The analysis, design, and development of formal methods and algorithms for the inconsistency, conflict resolution in the process of data/dnowledge bases and warehouses, integration, including the problems of different kinds of the fuzziness and of the uncertainty
- experimental implementations of new systems of Soft Computing.
- In all those domains the aim of creative synthesis of various approaches to Soft Computing is stressed.
Petr Hájek - CRCIM (UI)
Tel: +420 2 6605 3760