Holger Schwenk. Using boosting to improve a hybrid hmm/neural network speech recognizer. In ICASSP-99, 1999.  Home/Search Document Details and Download
Summary Related Articles Check |
This paper is cited in the following contexts:
Boosting Gaussian Mixtures In An LVCSR System - Zweig, Padmanabhan (2000) (3 citations) (Correct)
.... classifiers are then combined in such a way as to guarantee certain bounds on both training and testing error [2, 6] Boosting algorithms have been successfully applied to a wide variety of problems, including a recent application to boosting neural nets in a continuous speech recognition system [8]. One of the main advantages of boosting is that it is possible to automatically generate very long streams of classifiers anywhere from tens to thousands that usually produce better and better composite performance. Although boosting was originally presented as a method for combining ....
.... minimizing the true pseudoloss would require a gradient descent technique because changing the parameters associated with one class will affect all the others (see Equation 1) this simplification allows us to train models for each class independently and in parallel, and is also used in [8]. The parallelized algorithm is presented in Figure 1. The quantity Z t is a normalizing constant. The exact form of the parallel algorithm is motivated by the fact that with thousands of classes and tens of millions of examples, it is impossible to store D t (i; y) and therefore each time it is ....
Holger Schwenk. Using boosting to improve a hybrid hmm/neural network speech recognizer. In ICASSP-99, 1999.
Language Resources as by-Product of Evaluation: the.. - Patrick Paroubek Limsi (2000) (Correct)
....2. Combining to Improve In machine learning, it is well known that ensemble methods or committees of learning machines can often improve the performance of a system in comparison to a single learning machine. A very promising algorithm based on this principle now under investigation is Ada boost (Schwenk, 1999). In the same field, people have applied for a long time winner take all strategies to combine, inside the same system, the output of several basic processing units (Simpson, 1990) In the course of its evaluation program on speech recognition (S. 1998) NIST developed the ROVER (Recognizer ....
H. Schwenk. 1999. Using boosting to improve a hybrid hmm/neural network speech recognizer. In IEEE International Conference On Acoustics, Speech, and Signal Processing., Phoenix, USA, March.
A Neural Network For Classification With Incomplete Data - Morris (2000) (Correct)
.... likelihood based HMM system, using the same missing feature theory and the same method for detecting missing data [16] It is possible that the performance of the IDCN in this case could be improved by use of a more effective discriminative HMM training procedure, such as MCE [9] and or boosting [15]. When the position of missing data is not known, the IDCN offers a new approach to multi stream processing which should permit large numbers of feature streams to be combined with greatly reduced effort. This approach remains to be tested. Acknowledgements This work was supported by the EC OFES ....
Schwenk, H. (1999) "Using boosting to improve a hybrid HMM/neural network speech recogniser", Proc. ICASSP'99. pp.1009-1012.
Online articles have much greater impact More about CiteSeer.IST Add search form to your site Submit documents Feedback
CiteSeer.IST - Copyright Penn State and NEC