Neuromorphic Neural Network Parallelization on CUDA Compatible GPU for EEG Signal Classification

被引：4

作者：

Bako, Laszlo ^{[1
]}

Kolcsar, Arpad-Zoltan ^{[1
]}

Brassal, Sandor-Tihamer ^{[1
]}

Marton, Laszlo-Ferenc ^{[1
]}

Losonczi, Lajos ^{[2
]}

机构：

[1] Sapientia Hungarian Univ Transylvania, Dept Elect Engn, Targu Mures, Romania

[2] Lambda Commun Ltd, Targu Mures, Romania

来源：

2012 SIXTH UKSIM/AMSS EUROPEAN SYMPOSIUM ON COMPUTER MODELLING AND SIMULATION (EMS) | 2012年

关键词：

Spiking neural network; GPU; CUDA; parallelization; EEG; classification;

D O I：

10.1109/EMS.2012.87

中图分类号：

TP301 [理论、方法];

学科分类号：

081202 ;

摘要：

The purpose of the project described in this paper is to implement a Spiking Neural Network, on a CUDA driven Nvidia video-card, which can learn predefined samples on images presented as input data. With experimental EEG signals pre-processed using the Wavelet transform into an image set, it can learn to classify inputs into a certain category by applying a proprietary algorithm, presented in the paper. The implementation of the spiking neural network is done in CUDA C, with the use of the card's inner GPU. The GPU has the functionality to parallelize multiple tasks, which can enable the neural network to do fast calculations even with large amounts of data. The application can be controlled with a GUI, in which the user can modify the base parameters of the system, make tests, or it can train the system. Performance results are given in terms of computation speed and classification accuracy.

引用

页码：359 / 364

页数：6

共 14 条

[1] Unsupervised classification of complex clusters in networks of spiking neurons [J].