Automatically assigned DDC number:

Manually assigned DDC number: 00632

Number of references: 0

Title: International Journal of Neural Systems, Vol. ?, (Supp. 1995) 000--000 NOISY INFORMATION TRANSMISSION: A STABILITY ANALYSIS

Author:

Author:

Author:

Subject: Ro Campa,Paolo Del Giudice,Nestor Parga International Journal of Neural Systems, Vol. ?, (Supp. 1995) 000--000 NOISY INFORMATION TRANSMISSION: A STABILITY ANALYSIS

Description: Introduction The problem of learning the statistical properties of the environment has already received some attention 1 Gamma 7 (See also Ref. 8 for a review). For instance, for a noiseless channel the generalized Oja rule 3 produces synaptic couplings between an input and an output layer with M neurons that converge to values that span the same subspace as the M principal components. For a similar problem but including the effect of noise Linsker 5 has considered the infomax principle and, under some conditions, he exhibited a solution for the optimal couplings. This work however leaves many questions open about the behaviour of the network under different or more general conditions. We dropped some of these (e.g., translational invariance) and solve for the optimal couplings under different constraints. In this work we present the results for one kind of constraint, leavin

Contributor: The Pennsylvania State University CiteSeer Archives

Publisher: unknown

Date: 1998-03-05

Pubyear: unknown

Format: ps

Identifier: http://citeseer.ist.psu.edu/140822.html

Source: http://www.ft.uam.es/~parga/ARTICLES/CONFERENCES/elba94_linear.ps.gz

Language: en

Rights: unrestricted