| Paper: | MLSP-P3.11 | ||
| Session: | Pattern Recognition | ||
| Time: | Wednesday, May 17, 14:00 - 16:00 | ||
| Presentation: | Poster | ||
| Topic: | Machine Learning for Signal Processing: Signal detection, Pattern Recognition and Classification | ||
| Title: | Fast Incremental Techniques For Learning Production Rule Probabilities In Radar Electronic Support | ||
| Authors: | Guillaume Latombe, Éric Granger, École de Technologie Supérieure, Canada; Fred A. Dilkes, Defence Research and Development Canada, Canada | ||
| Abstract: | Although Stochastic Context-Free Grammars appear promising for recognition of radar emitters, and for estimation of their respective level of threat in radar Electronic Support systems, well-known techniques for learning their production rule probabilities are computationally demanding. In this paper, three fast incremental alternatives, called Graphical EM, Tree Scanning, and HOLA, are compared from several perspectives -- perplexity, estimation error, time and space complexity, and convergence time. Estimation of the execution time and storage requirements allows for the assessment of complexity, while computer simulation using a radar pulse data set allows to asses the other performance measures. Results indicate that Graphical EM and Tree Scanning may provide a greater level of accuracy than HOLA, whereas the computational efficiency may be orders of magnitude lower with HOLA. Furthermore, HOLA is an on-line technique that allows to reflect changes in the environment, by incremental learning of training sequences. | ||
©2018 Conference Management Services, Inc. -||- email: webmaster@icassp2006.org -||- Last updated Friday, August 17, 2012