A mesa rather than a peak? - a fitness landscape on weight space of an application using spiking neurons under rate coding

Author

Imada, Akira

Author_Institution

Brest State Tech. Univ.

fYear

2003

fDate

8-10 Sept. 2003

Firstpage

106

Lastpage

108

Abstract

We simulate an associative memory model using spiking neurons. Those models in which we specify interaction among McCulloch-Pitts neurons by a Hebbian-like learning algorithm, for example, already exist. The Hopfield model is one of these examples. Though we have still many unknown issues in the Hopfield model of associative memory, we have some inevitable drawbacks as well, such as small storage capacity. To overcome these drawbacks, and more importantly, to be more biologically plausible, we explore the model using spiking neurons

Keywords

Hebbian learning; content-addressable storage; encoding; evolutionary computation; neural nets; Hebbian-like learning algorithm; Hopfield model; McCulloch-Pitts neurons; associative memory model; downhill walk; evolutionary hill climbing; fitness landscape; rate coding; spiking neurons; storage capacity; synaptic weight space; Associative memory; Biological neural networks; Biological system modeling; Brain modeling; Differential equations; Face recognition; Hopfield neural networks; Humans; Neural networks; Neurons;

fLanguage

English

Publisher

ieee

Conference_Titel

Intelligent Data Acquisition and Advanced Computing Systems: Technology and Applications, 2003. Proceedings of the Second IEEE International Workshop on

Conference_Location

Lviv

Print_ISBN

0-7803-8138-6

Type

conf

DOI

10.1109/IDAACS.2003.1249527

Filename

1249527