RecNorm: Simultaneous Normalisation and Classification Applied to Speech Recognition

By John Scott Bridle, Stephen J. Cox. 1990

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Speech & Audio

Citations in the corpus

Learning Robust Global Representations by Penalizing Local Predictive Power (2019, background): “A broad set of papers have addressed various formulations of DA (Bridle and Cox, 1991; Ben-David et al., 2010a) dating in the ML and statistics literature to early works on covariate shift Shimodaira (2000) with antecedents classic econometrics work on sample selection bias (Heckman, 1977; Manski an”
Self-Challenging Improves Cross-Domain Generalization (2020, unknown): “To account for this discrepancy, technologies have been invented under the domain adaptation regime [2,3], where the goal is to train a model invariant to the distributional differences between the source domain (i.”
Adapting and controlling DNN-based speech synthesis using input codes (2017, methodology): “Following [23], we use backpropagation (BP) to minimise the mean square prediction error over a small amount of data uttered by the target speaker.”
Adapting Convolutional Neural Networks for Geographical Domain Shift (2019, background): “, adaptation of machine learning models so that they would work well on a target dataset different from the one they trained on (source dataset) has been considered in maching learning for a long time [5, 23, 35, 37], including theoretical results that connect domain adaptation with the theory of H∆”
Learning hidden unit contributions for unsupervised speaker adaptation of neural network acoustic models (2014, background): “Other approaches, related to the use of auxiliary features, are based on speaker codes [30, 31] in which a specific set of units for each speaker is optimised.”
Global optimization of a neural network-hidden Markov model hybrid (1991, methodology): “Insuch a system, the ANN adaptation represents a tuning of the feature space to the new speaker, whereas thetemporal model remains unchanged (see (Bridle & Cox 91) for a related speaker adaptation mechanism).”
When Networks Disagree: Ensemble Methods for Hybrid Neural Networks (1992, methodology): “…improve results in classiication and regression problems (Cooper, 1991; Reilly et al., 1988; Reilly et al., 1987; Scooeld et al., 1991; Baxt, 1992; Bridle and Cox, 1991; Buntine and Weigend, 1992; Hansen and Salamon, 1990; Intrator et al., 1992; Jacobs et al., 1991; Lincoln and Skrzypek, 1990;…”
Pattern recognition and neural networks (1996, background): “been widely suggested more recently, for example by Lincoln & Skrzypek (1990); Bridle & Cox (1991) ; Wolpert (1992); Xu et al. (1992); Perrone & Cooper (1993).”
Unsupervised Classifiers, Mutual Information and 'Phantom Targets' (1991, background): “The above phantom t.arget. rule has been re-derived for a larger class of networks [4], but the conditions for strict applicability are quite severe.”