Infobright Blog

Hello,

I spent last couple of days in Moscow, at the Higher School of Economics (HSE), where, together with Sergei Kuznetsov, we organized the first-ever international rough set conference in Russia. Besides participants from US, EU, China, India, and Japan, there was a strong group of Russian researchers and practitioners specialized in such areas as data mining, soft computing, semantic search and analytics, image recognition and processing, and many others.

The conference was relatively small but intensive. It started on Saturday with two workshops and two tutorials in the morning and the excellent opening keynote by Jiawei Han in the afternoon. I'm sure that most of you have heard about Jiawei's projects at the University of Illinois and perhaps had an occasion to read some of his articles. In Moscow, he talked about heterogeneous information networks - their construction and application to the web-based analysis of the growth of computer science. An information network is heterogeneous if its nodes represent the objects of different types (for example: papers, authors, conferences, sub-areas of science). Connections between nodes enable then to implement more powerful search and prediction methods than in case of homogeneous networks. (For example, isn't it nice to compare scientists by means of multiple aspects, such as attended conferences, co-authored and co-cited publications, or common research interests? - Yes, but then you need to work with all those types of nodes within the single network.) Surely, construction of heterogeneous networks requires mixed sources of information, which brings a lot interesting challenges, such as verifying inconsistencies and detecting misleading information, or modifying the standard ranking techniques to let them describe smaller clusters of more homogeneous objects.

Out of many other interesting presentations, let me briefly mention about the plenary talk by C. A. Murthy, who developed a brand new rough-set-based approach to optimizing ensembles of classifiers, performing far better than widely known machine learning techniques (see also his original paper applying this approach to webpage classification, as well as a more recent publication related to web services categorization and focused crawling), as well as the best paper award winner - Krzysztof Myszkorowski, who extended classical database axioms related to normal forms and functional dependencies towards data sets with values in form of intervals representing possibility distributions. (The paper is not yet available online, but let me refer you to the previous Krzysztof's publication.)

In summary, all important links, as well as information about other papers and presentations can be found at the conference homepage. I hope you will soon find there also some pictures taken during the sessions and the Moscow tour. Moscow is actually a very special city, with long and interesting history, lots of beautiful places and great atmosphere, so we were very fortunate to hold a rough set meeting here. Special thanks go to Sergei, who - in spite of his everyday responsibilities as the head of the School of Applied Mathematics and Information Science - managed to assembly the team of professors and students looking after the conference program and overall organization. I am returning to Poland today after the industry presentations but the accompanying events will last at HSE until Thursday. You may be sure that I will write about them as well!

Best greetings,

Dominik