Longbing Cao - Non-IIDness Learning in Big Data

Organized by: 

Yves Demazeau, équipe MAGMA


Professor Longbing Cao, Director of Advanced Analytics Institute, UTS

Professor Longbing Cao
Professor Longbing Cao

Dans le cadre de sa visite en tant que Prof. Invité UJF dans l’equipe MAGMA du 15 mai au 15 juillet, le Prof. Longbing Cao donnera la conférence suivante :


Longbing Cao is a professor of information technology at the Faculty of Engineering and IT, UTS ; and the founding Director of the UTS Advanced Analytics Institute. AAI is the largest research group in Australia focusing on advanced analytics, with broad collaborations with many major local and international organizations. Longbing was awarded PhD in computing sciences and PhD in intelligent sciences. Before joining UTS, Longbing had several years of research experience in Chinese Academy of Sciences, and working experiences in managing and leading industry and commercial projects in telecommunications, banking and publishing, as manager or chief technology officer. Besides general interest in areas such as data mining, machine learning, artificial intelligence, multi-agent systems and software engineering, Longbing has been initiating and now leading research in particular topics including behavior informatics and computing, non-iidness learning (including object relation learning and pattern relation learning), agent mining, and complex intelligent systems, and in particular enterprise applications of data mining and behavior informatics in the real world. In Australia, Longbing has solid links with broad-based major business, industry, vendor and government organizations, leading and managing many projects such as in social security, taxation, banking, telecommunication, capital market, insurance, public sector and airline business. During these exercises, Longbing fosters a strong research culture of conducting cutting-edge and applied research inspired by challenging but critical business and social problems, and forming a strong interaction and balance between high quality Research, high calibre analyst Education and high impact Development (so-called RED).

Big data is becoming a big thing for theoretical and technical innovation and for bigger business, smarter decisions and bigger value. This seminar first presents an overview of the current big data world, including concepts, state-of-the-art progress, challenges and opportunities. We then focus on discussing a recently emerging topic : Non-IIDness Learning, which handles two of fundamental challenges in big data analytics : heterogeneity and couplings. Heterogeneity and couplings in big data fundamentally challenge the classic IIDness foundation in statistics, data mining and machine learning etc. As a result, most of existing algorithms and approaches may not work for big data. The limitations of existing IIDness-based analysis, mining and learning are discussed, followed by exemplar techniques and case studies for object and pattern relation analysis, such as coupled clustering, coupled ensemble clustering, coupled behavior analysis, and relation analysis between terms, items, and patterns to tackle non-IIDness in the data.