Complexity-constrained feature selection for classification: 2007 DIGEST OF TECHNICAL PAPERS INTERNATIONAL CONFERENCE ON CONSUMER ELECTRONICS
2007 (English)In: IEEE ICCE, 2007, 9-10 p.Conference paper (Refereed)
Continuous monitoring of audio-visual context on mobile devices requires algorithms with gentle demands on computational resources. Existing feature selection strategies for classification do not account for the complexity associated with feature extraction. We present a complexity-constrained feature selection algorithm that is independent of the classifier architecture and demonstrate that it leads to superior feature sets if the allowed computational complexity is limited.
Place, publisher, year, edition, pages
2007. 9-10 p.
Computer and Information Science
IdentifiersURN: urn:nbn:se:kth:diva-36128DOI: 10.1109/ICCE.2007.341371ISI: 000245430200005ScopusID: 2-s2.0-34548026057OAI: oai:DiVA.org:kth-36128DiVA: diva2:430504
25th IEEE International Conference on Consumer Electronics Las Vegas, NV, JAN 10-14, 2007
QC 201107112011-07-112011-07-082011-07-11Bibliographically approved