Japan Advanced Institute of Science and Technology

Japan Advanced Institute of Science and Technology

JAIST Repository

https://dspace.jaist.ac.jp/

Title

ル

Author(s)

Citation

Issue Date

Type

Text version

URL

Rights

Description

underlying neuronal communication

Koutaro TORII

Scho ol ofInformation Science,

Japan Advanced Instituteof Scienceand Technology

July 10, 1997

Abstract

Theplasticityphenomenonofa neuronisconsideredasaleadingfactor formemorymanage-

ment and learning in a brain. However, investigation on this phenomenon hasbeensparse. In

order toinvestigate this subject, it is necessary to study the mechanism of information repre-

sentationand processing in a brain. In the present study, themechanisms related to synaptic

plasticity are thoroughly investigated and a scientic model for neural information processing

hasbeenproposed. The analyzed result isto be investigatedfordevelopinga learning metho d

related to information transfer mechanism. In the present study, apart from the exp erimental

results, amodel forsynapticplasticityisalso prop osed.

In order toexplore the transfer function of a nervous system, it is necessary to analyze the

results thathave b een achieved through electro-physiological experiments. Quantal analysis is

well knowfor these kind of data analysis. However, this method is very sensitive tonoise and

the precision is not go od. Hence, an attempt has been made to mo dify MEND (Maximum

Entropy Noise Deconvolution), which is considered to b e one of the best method for quantal

analysis. In the modied mo del, an extremely precise integration function is included for an

ecient utilization of the essence of the originalinformation that lies underneath thediscrete

experimental data. Compared tothe conventionalstatistical metho ds, the proposed improved

MENDgavebetterresults.

Presently,theinputinformationof aneuronisconsideredtobe theaveragenumb erofpulses

reaching a cell body. However, there are chances that information in the cell body is b eing

coded in the form of a rhythm or oscillation like factor. In the present approach, pulses with

dierent intervals are used for expressing information and the metho d has b een experimented

on rat cerebellum. The experimental resultsshowthat it isp ossible to expressinformation by

providing dierent stimulating patterns withan average frequency. The prop osed method has

been applied for the analysis of data achieved from the exp eriments on rat cereb ellum. The

analyzed resultis tob einvestigatedfordevelopinga learning metho d.

Key Words: quantal analysis, noise deconvolution, synaptic plasticity, tempo-

ral coding, hippocampus, cerebellum, patch-clamp