Departamento de Inform'atica
Download:
|
|
by S~ao Carlos Sp, Germano C. Vasconcelos, Michael C. Fairhurst, David L. Bisset
http://www.di.ufpe.br/~compint/./papers/IIsbrn.ps.gz
Add To MetaCart
Abstract:
Multilayer perceptron networks (MLP) are investigated with respect to the problem of the detection of spurious patterns. A novel mechanism is proposed based on the ideas of bootstrapping that when incorporated into the standard MLP provides the network with the ability to continuously modifying its responses across the input space. The mechanism makes use of the outputs given by the network itself during its recall phase to improve performance in the rejection of spurious inputs, through a scheme of reinforcement of the classification decisions taken by the network. A example of the 1
Citations
| 2140 | Learning Internal Representations by Error Propagation – Rumelhart, Hinton, et al. - 1986 |
| 133 | Adaptive signal processing – Widrow, Stearns - 1985 |
| 60 | Novelty Detection and Neural Network Validation – Bishop - 1994 |
| 27 | Punishlreward: Learning with a critic in adaptive threshold systems – Widrow, Gupta, et al. - 1973 |
| 27 | Inversion of multilayer nets – Linden, Kindermann - 1989 |
| 21 | Techniques for adaptive equalization of digital communication systems – Lucky - 1966 |
| 5 | Investigating feedforward neural networks with respect to the rejection of spurious patterns – Vasconcelos, Fairhurst, et al. - 1995 |
| 2 | Efficient detection of spurious inputs to improve the robustness of MLP networks in practical applications – Vasconcelos, Fairhurst, et al. - 1995 |
