N. Intrator and S. Edelman. Making a low-dimensional representation suitable for diverse tasks. Connection Science, 8(2):205--224, 1997. Home/Search Document Details and Download
Summary Related Articles Check |
This paper is cited in the following contexts:
. Model Selection of Combined Neural Nets for Speech.. - Summary The Problem (Correct)
....experiments that allows transfer of the perceptual classification ability to the acoustic space of a new user of the system. Acoustic normalization methods are indeed instances of user adaptation methodologies, and can be seen as an example in the use of transfer of perceptual classification [37, 27], as learning an appropriate mapping between acoustic feature spaces facilitates generalization in speech recognition. The application of bootstrap procedures to model selection in combined neural networks seems to provide an interesting tuning method in automatic speech recognition problems. The ....
N. Intrator and S. Edelman. Making a low-dimensional representation suitable for diverse tasks. Connection Science, 8(2):205--224, 1997.
Prototype Selection for Composite Nearest Neighbor Classifiers - Skalak (1997) (10 citations) (Correct)
....are created. 24 Additional support for a small number of component classifiers comes from different source psychological studies of human perception. In particular, the human visual system apparently performs as though the stimuli [exist] in a low dimensional metric psychological space [Intrator and Edelman, 1996, p.4] Intrator and Edelman use this observation and other evidence from experiments with human vision to argue that a low dimensional representation supports the transfer of skill across tasks. For our composite classifiers, fewer component classifiers entails a smaller number of features in ....
Intrator, N. and Edelman, S. 1996. Making a low-dimensional representation suitable for diverse tasks. Connection Science, to appear.
Improving Classification via Reconstruction - Stainvas, Intrator (2000) Self-citation (Intrator) (Correct)
....and in order to control this variance, innovative bias constraints should be used Stainvas et al. 3 (Geman et al. 1992) One way is to construct e#cient low dimensional representations which are su#cient for the classification task. For example, this can be achieved by multiple class constraints (Intrator and Edelman, 1996). In this paper, we continue this line of thought and study the e#ect of regularization in the form of reconstruction constraint on the resulting classifier. We further study if reconstruction constraints can replace, or should be added to weight decay (WD) Reconstruction constraints have been ....
....Neural Network Ensembles Another way to assess the e#ect of regularization constraints is by combining regularized networks into ensembles. It is well known that an ensemble of experts is capable of improving the performance of single experts (Wolpert, 1992; Krogh and Vedelsby, 1995; Raviv and Intrator, 1996). There are two main questions to be addressed when constructing ensembles: i) how to evaluate an ensemble classification prediction from predictions of its members and (ii) which networks to combine. There are di#erent ways to evaluate an ensemble classification prediction. The first, is using ....
[Article contains additional citation context not shown here]
Intrator, N. and Edelman, S. (1996). Making a low-dimensional representation suitable for diverse tasks. Connection Science, Special issue on Reuse of Neural Networks Through Transfer, 8(2):205--224. Also in Learning to Learn, S. Thrun and L. Pratt ed., Kluwer press.
Improving Classification via Reconstruction - Stainvas, Intrator, al. (2000) Self-citation (Intrator) (Correct)
....high variance and in order to control this variance, innovative bias constraints should be used (Geman et al. 1992) One way is to construct efficient low dimensional representations which are sufficient for the classification task. For example, this can be achieved by multiple class constraints (Intrator and Edelman, 1996). In this paper, we continue this line of thought and study the effect of regularization in the form of reconstruction constraint on the resulting classifier. We further study if reconstruction constraints can replace, or should be added to weight decay (WD) Reconstruction constraints have been ....
....Neural Network Ensembles Another way to assess the effect of regularization constraints is by combining regularized networks into ensembles. It is well known that an ensemble of experts is capable of improving the performance of single experts (Wolpert, 1992; Krogh and Vedelsby, 1995; Raviv and Intrator, 1996). There are two main questions to be addressed when constructing ensembles: i) how to evaluate an ensemble classification prediction from predictions of its members and (ii) which networks to combine. There are different ways to evaluate an ensemble classification prediction. The first, is using ....
[Article contains additional citation context not shown here]
Intrator, N. and Edelman, S. (1996). Making a low-dimensional representation suitable for diverse tasks. Connection Science, Special issue on Reuse of Neural Networks Through Transfer, 8(2):205--224. Also in Learning to Learn, S. Thrun and L. Pratt ed., Kluwer press.
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