Abstract | ||
---|---|---|
Differentiation between the nodes of a competitive learning net- work is conventionally achieved through competition on the ba- sis of neural activity. Simple inhibitory mechanisms are limited to sparse representations, while decorrelation and factorization schemes that support distributed representations are computa- tionally unattractive. By letting neural plasticity mediate the com- petitive interaction instead, we obtain diffuse, nonadaptive alter- natives for fully distributed representations. We use this tech- nique to simplify and improve our binary information gain op- timization algorithm for feature extraction (Schraudolph and Se- jnowski, 1993); the same approach could be used to improve other learning algorithms. |
Year | Venue | Keywords |
---|---|---|
1994 | NIPS | neural plasticity,feature extraction,competitive learning,information gain,sparse representation |
Field | DocType | Citations |
Competitive learning,Decorrelation,Computer science,Binary information,Neural activity,Feature extraction,Factorization,Artificial intelligence,Optimization algorithm,Machine learning,Plasticity | Conference | 0 |
PageRank | References | Authors |
0.34 | 4 | 2 |
Name | Order | Citations | PageRank |
---|---|---|---|
Nicol N. Schraudolph | 1 | 1185 | 164.26 |
Terrence J. Sejnowski | 2 | 8278 | 2135.10 |