|
DukeSpace at Duke University >
University Archives >
Pratt School of Engineering >
Technical Reports >
Please use this identifier to cite or link to this item:
http://hdl.handle.net/10161/1286
|
| Title: | Sequential anomaly detection in the presence of noise and limited feedback |
| Authors: | Raginsky, Maxim
Willett, Rebecca |
| Publication Date: | 13-Aug-2009 |
| Abstract: | Recent work has established the efficacy of using online convex programming methods on exponential families in the context of sequential probability assignment. This paper describes methods which build
upon that framework to handle noisy observations. Furthermore, the problem of detecting anomalous (i.e. rare) events by using the sequential probability assignments and limited feedback is presented. |
| Keywords: | Filtering
universal prediction
individual sequences
anomaly detection |
| URI: | http://hdl.handle.net/10161/1286 |
| Series/Report no.: | ECE-2009-01 |
| Appears in Collections: | Technical Reports
|
All items in DukeSpace are protected by copyright, with all rights reserved.
|
DukeSpace
