Magazine article AI Magazine

The Distributed Data-Mining Workshop

Magazine article AI Magazine

The Distributed Data-Mining Workshop

Article excerpt

The Workshop on Distributed Data Mining (DDM), held at the Fourth International Conference on Knowledge Discovery and Data Mining, brought interested researchers and practitioners together and created an environment for crystallizing the fast-growing field of DDM. The workshop focused on the state-of-the-art DDM algorithms, systems, and application-related issues. Approximately 40 participants attended the workshop. The workshop had 13 presentations, including 3 invited talks.

The workshop began with a visionary invited talk on the possible role of mobile agents in DDM by George Cybenko of Dartmouth University. His talk described some of the existing mobile agent-based systems and pointed out the strengths of distributed data analysis using mobile agents. This talk was followed by the general session. Yike Guo and Janias Sutiwaraphun (both of Imperial College, United Kingdom) proposed distributed knowledge probing that uses meta learning for hierarchical supervised learning. Hillol Kargupta, Erik Johnson, Byung-Hoon Park, Daryl Hershberger (all of Washington State University), Eleonara Riva Sanseverino, and Luisa Di Silvestre (both of University of Palermo, Italy) proposed a new approach called collective data mining from distributed data with vertically partitioned feature space. They pointed out that naive local analysis of data can be misleading when features are distributed, and distributed representation of functions using a set of orthogonal basis functions can effectively be used for accurate DDM. Vincent Ng and David Cheung (both of Hong Kong Polytechnic University, China) presented new techniques for learning distributed associations under constraints. Philip Chan (Brigham Young University) and Salvatore Stolfo (Columbia University) presented the effects of class distribution in the training set on performance and proposed different techniques for measuring performance. Lawrence Hall, Nitesh Chawla, and Kevin W. Bowyer (all of University of South Florida) presented the results of their approach for combining decision trees in parallel. Andreas Prodromidis (Columbia University) and Stolfo suggested different techniques for pruning metaclassifiers to make the resultant system less complex. Abdula Tansel (City University of New York) and Necip Ayan (Bilkent University) discussed techniques for discovering association rules in temporal data.

The afternoon session was initiated by a stimulating invited talk by Foster Provost (Bell Atlantic Science and Technology) on the scalability issue of DDM. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.