Constrained corrective training for continuous parameter system

作者: Dimitri Kanevsky

DOI:

关键词:

摘要: A method is provided for training a statistical pattern recognition decoder on new data while preserving its accuracy of old, previously learned data. Previously are represented as constrained equations that define domain (T) in space parameters (K) the decoder. Some part feasible point domain. procedure reformulated optimization objective functions over Finally, solved. This ensures preserved during iterative steps. While an exemplary speech discussed, inventive also suited to other problems such as, example, handwriting recognition, image machine translation, or natural language processing.

参考文章(6)
John G. Proakis, John R. Deller, John H. Hansen, Discrete-Time Processing of Speech Signals ,(1993)
D. Kanevsky, A generalization of the Baum algorithm to functions on non-linear manifolds international conference on acoustics, speech, and signal processing. ,vol. 1, pp. 473- 476 ,(1995) , 10.1109/ICASSP.1995.479631
Lalit R. Bahl, Frederick Jelinek, Robert L. Mercer, A Maximum Likelihood Approach to Continuous Speech Recognition IEEE Transactions on Pattern Analysis and Machine Intelligence. ,vol. PAMI-5, pp. 179- 190 ,(1983) , 10.1109/TPAMI.1983.4767370
P.S. Gopalakrishnan, D. Kanevsky, A. Nadas, D. Nahamoo, An inequality for rational functions with applications to some statistical estimation problems IEEE Transactions on Information Theory. ,vol. 37, pp. 107- 113 ,(1991) , 10.1109/18.61108