Bandwidth-efficient similarity search on time series data in an machine-to-machine environment
Mi-Yen Yeh (Academia Sinica, Taiwan)
NICTA SML SEMINAR Machine Learning Research Group SeminarDATE: 2013-04-18
TIME: 11:45:00 - 12:15:00
LOCATION: NICTA - 7 London Circuit
CONTACT: JavaScript must be enabled to display this email address.
ABSTRACT:
A machine-to-machine (M2M) environment refers to a network of devices capturing various readings. When performing any computing tasks in an M2M environment, transmission cost including the energy and bandwidth consumption is always one of the most important concerns. This talk will describe our work to do ad hoc similarity search on time series in an M2M environment when available bandwidth is limited. We propose a unified framework to handle both k-nearest and k-furthest neighbor (kNN and kFN) queries in a distributed environment, while significantly reducing the bandwidth consumption without causing any false dismissals. More important, we can handle not only the kNN and kFN to one reference time series but also to a reference set comprising multiple ones. By exploiting the wavelet transformation together with the novel distance bound designs, our method obtains the exact results of kNN and kFN queries in a bandwidth-efficient manner. Analytical and empirical studies showing the significant bandwidth saving of our proposed method will also be described in the talk.
BIO:
Mi-Yen Yeh is currently Assistant Research Fellow (equivalent to tenure-track Assistant Professor) of Institute of Information Science at Academia Sinica, Taiwan. She received her Ph.D. degree in Electrical Engineering from National Taiwan University, Taiwan in 2009. Her main research area is on data mining and databases, with a specific focus on streaming time series analysis, trajectory mining, and social network analysis. She received the best paper award (in system software and security) of the 28th annual ACM Symposium on Applied Computing (SAC 2013), Distinguished Postdoctoral Fellowship in Academia Sinica, and Research Exploration Award in Pan Wen Yuan Foundation.
