In classification problems the most commonly used neural network is probably the multilayer perceptron network (MLPN). The probabilistic neural network (PNN) is a possible alternative to the MLPN. The PNN is based on the Bayesian approach and a non-parametric estimation of the probability density fu
Geometrical Aspects of Discrimination by Multilayer Perceptrons
β Scribed by Salvatore Ingrassia
- Publisher
- Elsevier Science
- Year
- 1999
- Tongue
- English
- Weight
- 107 KB
- Volume
- 68
- Category
- Article
- ISSN
- 0047-259X
No coin nor oath required. For personal study only.
β¦ Synopsis
We investigate some geometrical aspects of the discriminant functions of the kind
for suitable constants a k , c k where { is a sigmoidal transformation. This function is realized by a multilayer perceptron with one hidden layer. These results are applied in the analysis of the discriminating power of f p . In particular, we prove that the class of finite populations 0 1 and 0 2 that can be distinguished by f p is monotonically increasing in p and we give a minimal sufficient p leading to a complete separation between 0 1 and 0 2 .
π SIMILAR VOLUMES
Multilayer perceptrons can compute arbitrary dichotomies of a set of \(N\) points of \([0,1]^{d}\). The minimal size of such networks was studied by Baum (1988, J. Complexity 4, 193-215) using the parameter \(N\). In this paper, we show that this question can be addressed using another parameter, th
The immobilization of the globular protein β£-1-acid glycoprotein (AGP) onto silica gel led to the commercial availability of an AGP column, which has a high enantioselectivity. The enantioselectivity of AGP columns has been demonstrated in numerous applications. Due to potential AGP structural chang
We thank the EPSRC for support (EP/E030122), the Cardiff Institute of Tissue Engineering and Repair (CITER), and JREI (JR99BAPAEQ) for a 20 processor Compaq SC cluster at the Rutherford Appleton Laboratory.