Foreword

Scalability is one of the main problems practitioners have to cope with when grasping a real-world application in data management or information analysis. The size of databases and data warehouses, associated with incompleteness of information and missing values has been a major difficulty from the early beginning of their studies. Modern digital devices, Internet possibilities, and distributed networks are among the most powerful means of storing, retrieving, and sharing information. The amounts of documents and data available for the users are continuously increasing, whatever their nature may be: text, video, music, images, multimedia, Web. The ways to access these documents and data are also diverse: exchanges within communities, social networks and peer to peer communications have increased the complexity of transfers from data repositories to users.

To increase the efficiency of existing algorithms is a necessity. Dimension reduction or dynamic treatment of data avoiding their storage is for instance a solution to large scale learning systems. Moreover, alternative approaches to classic information retrieval, knowledge discovery and data analysis need to be created, in order to cope with the complexity of the problem to solve, due to the size, the heterogeneity, the incompleteness of data and their access paths. Thinking differently is also a necessity since classic statistics or machine learning methods have their limits. System science provides interesting paradigms for the handling of complex systems, always taking the user into account, in a holistic involvement of all components of the system. Active learning involving the user is for example a solution to the difficulty of using supervised learning in huge training sets. Another lesson from systems science is the exploitation of synergies between components of the system, and this capacity is well understood in the complementarity between medias, for instance between text and image.

Fuzzy knowledge representation and logic are among the efficient tools for the management of complex systems, since they bring solutions to the incompleteness, inaccuracy and uncertainty, inherent to large scale and heterogeneous information reservoirs, taking into account synthetic descriptions of isolated elements and reducing individual treatments. Providing an interface between numerical data representations by computers and symbolic representations well understood by humans, fuzzy logic fills in the gap between technological needs and usability requirements. Concepts such as fuzzy categories, fuzzy quantifiers, fuzzy prototypes, fuzzy aggregation methods, fuzzy learning algorithms, fuzzy databases, and fuzzy graphs have proved their utility in the construction of scalable algorithms.

The present book is certainly of particular interest for the diversity of addressed topics, covering a large spectrum in scalability management. Anne Laurent and Marie-Jeanne Lesot are experts in theoretical and methodological study of fuzzy techniques, and they have moreover coped with various real world large-scale problems. The group of experts they have gathered to prepare this volume is unquestionably qualified to provide solutions to researchers and practitioners in search of efficient algorithms and models for complex and large dataset management and analysis.

Scalability is understood in this book from several points of view. The first one is the size of available data implying difficulties in their tractability, with regard to memory size or computation time. This aspect is strongly related to the complexity of involved algorithms.

The second point of view regards the form of the algorithm results and the capability of human users to understand and grasp these results, through summaries and visualization solutions. This aspect is more related to a cognitive framework.

The scalability of knowledge representation is at the crossroads of these points of view, dealing with ontologies or formal languages, as well as a variety of concepts in a fuzzy setting.

The classic scalability problem in hardware is another point of view, revisited here in the light of modern electronic solutions and fuzzy computation.

This book deals with all these aspects under a fuzzy logic based perspective. A sample of applications is also presented as a showcase, pointing out the efficiency of fuzzy approaches to the construction of scalable algorithms. Potential applications of such approaches go far beyond the domains tackled here and this book opens the door to a vast spectrum of forthcoming works.

Bernadette Bouchon-Meunier

LIP6 / UPMC / CNRS, France

Bernadette Bouchon-Meunier is the head of the department of Databases and Machine Learning in the Computer Science Laboratory of the University Paris 6 (LIP6). Graduate from the Ecole Normale Superieure at Cachan, she received the degrees of B.S. in Mathematics and Computer Science, Ph.D. in Applied Mathematics and D. Sc. in Computer Science from the University of Pierre and Marie Curie. Editor-in-Chief of the International Journal of Uncertainty, Fuzziness and Knowledge-based Systems, she is a co-founder and co-executive director of the International Conference on Information Processing and Management of Uncertainty in Knowledge-based Systems (IPMU) held every other year since 1986. She is an IEEE senior member and chair of the IEEE French Chapter on Computational Intelligence.. Her present research interests include approximate and similarity-based reasoning, as well as the application of fuzzy logic and machine learning techniques to decision-making, data mining, risk forecasting, information retrieval and user modeling.