Research for Neuron Classification Based on Support Vector Machine

Author

Fengqing, Han ; Jie, Zeng

Author_Institution

Sch. of Sci., Chongqing Jiaotong Univ., Chongqing, China

fYear

2012

fDate

July 31 2012-Aug. 2 2012

Firstpage

646

Lastpage

649

Abstract

In this paper, a new method is proposed for neurons classifying based on its spatial structure. The part of neuron is geometrically similar to the whole. Neurons can be regarded as fractal. Different types of neurons fill with different levels in space. So, their fractal dimensions are also different. First, fractal dimensions are calculated for neurons. Then the other 16 spatial structure indicators are added in the classifier. There are 44 neurons as the training samples to train Support Vector Machine and other 20 neurons as the test samples. Experiments show that the correct classification rate is almost over 70% for many cases. It provides a new method to classify neurons.

Keywords

fractals; geometry; neural nets; pattern classification; support vector machines; correct classification rate; fractal dimensions; geometry; neuron classification; spatial structure; support vector machine; training samples; Animals; Fractals; Kernel; Mathematical model; Neurons; Support vector machines; Training; classification; fractal geometry; spatial structure; support vector machine;

fLanguage

English

Publisher

ieee

Conference_Titel

Digital Manufacturing and Automation (ICDMA), 2012 Third International Conference on

Conference_Location

GuiLin

Print_ISBN

978-1-4673-2217-1

Type

conf

DOI

10.1109/ICDMA.2012.153

Filename

6298600