Naslov
Stimulus preprocessing for holographic neural networks

Autori
Manger, Robert ; Plantamura, Vito Leonardo ; Souček, Branko

Vrsta, podvrsta i kategorija rada
Poglavlja u knjigama, znanstveni

Knjiga
Frontier Decision Support Concepts

Urednik/ci
Plantamura, Vito Leonardo ; Souček, Branko ; Visaggio, Giuseppe

Izdavač
John Wiley & Sons

Grad
New York (NY)

Godina
1994

Raspon stranica
79-90

ISBN
0-471-59256-0

Ključne riječi
artificial neural networks, holographic neural technology, preprocessing

Sažetak
There are two types of preprocessing which are of fundamental importance for holographic neural networks. The first type is stimulus symmetrization, which assures reasonable accuracy in reproducing the learned stimulus-response associations. The second type is stimulus expansion, which increases the number of stimulus-response associations that can accurately be encoded into a holographic neuron. In this chapter we review the standard preprocessing methods used for holographic neural networks, and also propose two new methods, i.e. one for each of the two types of preprocessing. Some situations are described where these new methods are useful, and some experimental results are listed. We believe that our proposals greatly enhance the applicability of holographic networks.

Izvorni jezik
Engleski

Znanstvena područja
Matematika, Računarstvo

POVEZANOST RADA