Vassilis Christophides (Université de Crète – Inria Paris) s’est rendu à l’IRT SystemX le 14 mars pour pour animer un séminaire sur le thème « Detecting Outliers in IoT Data Streams ».
Résumé (en anglais) :
IoT-based systems typically collect data continuously produced by a large number of devices (aka ìthingsî) and analyze them in (sub-)seconds in order to act promptly in different contexts (personal healthcare, cyber-physical security, industrial plant monitoring, etc.). A critical aspect for gaining insights from IoT data is to control their quality, i.e., check whether data exhibit various forms of impurities as for example outliers. An outlier data point maybe the outcome of an error in the data collection or sensing process (e.g., due to a problem with measuring device or the presence of noise) or of a change in the data generation process per se (e.g., unobserved data patterns, events).
In this talk, we will describe the requirements of IoT data analytics and present several algorithms for exact or approximate outlier detection. In particular, we will focus on semi-supervised and unsupervised methods for detecting outliers in batch and stream data processing. We will explain the definition ìoutlyingnessî, i.e., why outlier data differ from the other data, in distance or density-based algorithms along with their computational and communication requirements. We conclude with a number of open research issues for detecting outliers in collections of IoT data streams that might be geographically distributed in edge and fog architectures.
Biographie (en anglais) :
Vassilis Christophides is a Professor of Computer Science at the University of Crete. He has been appointed in 2015 to an advanced research position at Inria-Paris. Previously, he worked as a Distinguished Scientist at Technicolor, R&I Center in Paris. His main research interests include Databases and Web Information Systems, as well as Big Data Processing. His current research work at INRIA focuses on Network measurements and IoT data analytics. He has published over 130 articles in high-quality international conferences and journals. He has been scientific coordinator of a number of research projects funded by the European Union and the Greek State. He has received the 2004 SIGMOD Test of Time Award and 2 Best Paper Awards (ISWC 2003 and 2007). He served as General Chair of the EDBT/ICDT Conference in 2014 and as Area Chair for the ICDE ìSemi-structured, Web, and Linked Data Managementî track in 2016. He has also co-authored a book on entity resolution in the Web of data.
Pour aller plus loin :