TY - GEN
T1 - Enhanced Image Classification through Customized Convolutional Spiking Neural Network
AU - Saini, Ashok Kumar
AU - Kumar, Rajesh
AU - Gehlot, Naveen
AU - Verma, Seema
N1 - Publisher Copyright:
© 2024 IEEE.
PY - 2024
Y1 - 2024
N2 - Spiking Neural Networks (SNNs) are deemed to provide biological realism. Also, it has more computational power than Artificial Neural Networks (ANNs) due to its utilization of spikes for information transmission and encoding. However, their shallow structures impose structural limitations, restricting the feature extraction capabilities of conventional SNNs. This study aims to improve the feature extraction capability of SNNs by leveraging the proficient feature extraction skills of Convolutional Neural Networks (CNNs). Our proposed model, Customized Convolutional Spiking Neural Network (CCSNN), combines CNN for feature learning with SNNs for cognitive skills. On the Digit-MNIST, Fashion-MNIST, and Letter-MNIST datasets, CCSNN surpasses previous models using fewer neurons and less training data, enhancing the biological realism of image classification models. In this study, CCSNN achieved impressive results on the Digit-MNIST, Fashion-MNIST, and Letter-MNIST datasets, with accuracies of 99.10%, 91.80%, and 99.30%, respectively, compared to conventional SNN.
AB - Spiking Neural Networks (SNNs) are deemed to provide biological realism. Also, it has more computational power than Artificial Neural Networks (ANNs) due to its utilization of spikes for information transmission and encoding. However, their shallow structures impose structural limitations, restricting the feature extraction capabilities of conventional SNNs. This study aims to improve the feature extraction capability of SNNs by leveraging the proficient feature extraction skills of Convolutional Neural Networks (CNNs). Our proposed model, Customized Convolutional Spiking Neural Network (CCSNN), combines CNN for feature learning with SNNs for cognitive skills. On the Digit-MNIST, Fashion-MNIST, and Letter-MNIST datasets, CCSNN surpasses previous models using fewer neurons and less training data, enhancing the biological realism of image classification models. In this study, CCSNN achieved impressive results on the Digit-MNIST, Fashion-MNIST, and Letter-MNIST datasets, with accuracies of 99.10%, 91.80%, and 99.30%, respectively, compared to conventional SNN.
KW - Classification
KW - Convolutional Neural Network (CNN)
KW - MNIST
KW - Spiking Neural Network (SNN)
UR - http://www.scopus.com/inward/record.url?scp=85211911312&partnerID=8YFLogxK
U2 - 10.1109/PICET60765.2024.10716183
DO - 10.1109/PICET60765.2024.10716183
M3 - Conference contribution
AN - SCOPUS:85211911312
T3 - 2024 Parul International Conference on Engineering and Technology, PICET 2024
BT - 2024 Parul International Conference on Engineering and Technology, PICET 2024
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 6th Parul International Conference on Engineering and Technology, PICET 2024
Y2 - 3 May 2024 through 4 May 2024
ER -