In the previous years we have conducted research in the areas of databases and data mining. More specifically in unsupervised learning (clustering algorithms and validity measures), advanced data management and indexing (P2P systems, distributed indexing, distributed dimensionality reduction), text mining (word disambiguation for classification, introduced the Graph of Words approach) and ranking algorithms (temporal extensions to PageRank).
More recently, we worked in large scale graph mining (degeneracy based community detection and evaluation), text mining and retrieval for web advertising/marketing and recommendations.
Current research interests and work include the areas:
- machine learning for graphs (graph kernels, embedding methods, deep learning for graph classification, large scale community detection) with applications in fraud detection
- NLP and text mining (Graph of Words, Deep learning for text classification, summarization and keyword extraction) applications
- decision making methods, in particular: mathematical programming (mixed integer linear and nonlinear programming), combinatorial optimization, global optimization, graph theory. We are interested both in methodology and applications, with a special focus on applications in energy optimization and computational geometry.
- event and anomaly detection in data streams and time series (applications in text streams, sensory data, personalized medicine)
- structured output prediction (multi-label classification, multi-output and sequential/dynamical models, probabilistic models and neural networks)
- reinforcement learning (Bayesian models, and deep learning)>
The DaSciM team members have supervised fifteen completed Ph.D. theses and published chapters in books and encyclopedias, two international books and more than a 250 papers in international refereed journals and conferences. Also we have co-authored three patents and attracted significant R&D funding including national and international governmental/industrial sources. Members of our team have received the ERCIM, Marie Curie, and Google fellowships. Our team has co-organized the ECML PKDD 2011 conference in Athens, ECML/PKDD 2017 and participates in the senior organization of different AI and Data mining related events (AAAI, IJCAI).
Moreover our group has a long experience in real-world R&D projects in the area of Large Scale Data/Text/time series Mining. Currently we maintain collaborations with industrial partners (including AIRBUS, Google, BNP, Tencent, Tradelab) working on machine learning projects.
Professor Vazirgiannis leads the X/AXA Data Science Chair.
- our position paper entitled “Search in BigData2 – When Big Text meets Big Graph”, presented on the World Summit on Big Data and Organization Design, organized by IBM, Organizational Design Community, and the Interdisciplinary Center for Organizational Architecture/ Aarhus University, on May 16-17, 2013 in Paris.
- the ERCIM White Paper on Big Data Analytics here.
Another relevant research topic is decision making methods, in particular: mathematical programming (mixed integer linear and nonlinear programming), combinatorial optimization, global optimization, graph theory. We are interested both in methodology and applications, with a special focus on applications in energy optimization and computational geometry.
Visit our selected publications page.
mvazirg ~ lix.polytechnique.fr