by Horst Bischof
Abstract:
In order to apply neural networks to large scale, real world tasks, several obstacles have to be overcome. One main deficiency is long learning time. A closer look at the brain reveals that the topology of the brain is considerably different from current neural network models. We show that modular and hierarchical toplogies (also common in the brain) offer a potential solution to speed up learning. To build modular and hierarchical network topologies, knowledge from conventional
Reference:
Modular, Hierarchical, and Geometrical Neural Networks (Horst Bischof), Technical report, PRIP, TU Wien, 1991.
Bibtex Entry:
@TechReport{TR009,
author = "Horst Bischof",
institution = "PRIP, TU Wien",
number = "PRIP-TR-009",
title = "Modular, {H}ierarchical, and {G}eometrical {N}eural
{N}etworks",
year = "1991",
abstract = "In order to apply neural networks to large scale,
real world tasks, several obstacles have to be
overcome. One main deficiency is long learning
time. A closer look at the brain reveals that the
topology of the brain is considerably different from
current neural network models. We show that modular
and hierarchical toplogies (also common in the
brain) offer a potential solution to speed up
learning. To build modular and hierarchical network
topologies, knowledge from conventional",
url = "https://www.prip.tuwien.ac.at/pripfiles/trs/tr9.pdf",
}