The original paper is in English. Non-English content has been machine-translated and may contain typographical errors or mistranslations. ex. Some numerals are expressed as "XNUMX".
Copyrights notice
The original paper is in English. Non-English content has been machine-translated and may contain typographical errors or mistranslations. Copyrights notice
심층 신경망의 지속적인 혁신과 함께 생물학적 뇌 시냅스와 더욱 유사한 스파이킹 신경망(SNN)이 낮은 전력 소비로 인해 주목을 받고 있습니다. 인공 신경망(ANN)과 달리 연속 데이터 값의 경우 인코딩 프로세스를 사용하여 값을 스파이크 트레인으로 변환하여 SNN의 성능을 억제해야 합니다. 이러한 품질 저하를 방지하려면 들어오는 아날로그 신호를 인코딩 프로세스 이전에 조절해야 합니다. 이는 또한 생명체에서도 실현됩니다. 예를 들어 인간의 기저막은 기계적으로 푸리에 변환을 수행합니다. 이를 위해 ANN과 SNN을 결합하여 해당 성능을 향상시키는 ANN-to-SNN 하이브리드 신경망(HNN)을 구축합니다. 이러한 성능과 견고성을 검증하기 위해 훈련 및 인코딩 방법이 변경되는 다양한 분류 작업에 MNIST 및 CIFAR-10 이미지 데이터 세트가 사용됩니다. 또한, 인공 레이어와 스파이킹 레이어 각각의 인코딩 방식을 고려하여 동시 학습과 분리 학습 방법을 제시한다. 우리는 스파이킹 레이어를 희생하여 인공 레이어 수를 늘리면 HNN 성능이 향상된다는 것을 발견했습니다. MNIST와 같은 간단한 데이터 세트의 경우 중복 코딩 및 별도의 학습을 사용하여 ANN과 유사한 성능을 얻을 수 있습니다. 그러나 더 복잡한 작업의 경우 가우스 코딩과 동시 학습을 사용하면 HNN의 정확도를 향상시키는 동시에 전력 소비를 낮추는 것으로 나타났습니다.
Naoya MURAMATSU
University of Cape Town
Hai-Tao YU
University of Tsukuba
Tetsuji SATOH
University of Tsukuba
The copyright of the original papers published on this site belongs to IEICE. Unauthorized use of the original or translated papers is prohibited. See IEICE Provisions on Copyright for details.
부
Naoya MURAMATSU, Hai-Tao YU, Tetsuji SATOH, "Combining Spiking Neural Networks with Artificial Neural Networks for Enhanced Image Classification" in IEICE TRANSACTIONS on Information,
vol. E106-D, no. 2, pp. 252-261, February 2023, doi: 10.1587/transinf.2021EDP7237.
Abstract: With the continued innovation of deep neural networks, spiking neural networks (SNNs) that more closely resemble biological brain synapses have attracted attention because of their low power consumption. Unlike artificial neural networks (ANNs), for continuous data values, they must employ an encoding process to convert the values to spike trains, suppressing the SNN's performance. To avoid this degradation, the incoming analog signal must be regulated prior to the encoding process, which is also realized in living things eg, the basement membranes of humans mechanically perform the Fourier transform. To this end, we combine an ANN and an SNN to build ANN-to-SNN hybrid neural networks (HNNs) that improve the concerned performance. To qualify this performance and robustness, MNIST and CIFAR-10 image datasets are used for various classification tasks in which the training and encoding methods changes. In addition, we present simultaneous and separate training methods for the artificial and spiking layers, considering the encoding methods of each. We find that increasing the number of artificial layers at the expense of spiking layers improves the HNN performance. For straightforward datasets such as MNIST, similar performances as ANN's are achieved by using duplicate coding and separate learning. However, for more complex tasks, the use of Gaussian coding and simultaneous learning is found to improve the accuracy of the HNN while lower power consumption.
URL: https://global.ieice.org/en_transactions/information/10.1587/transinf.2021EDP7237/_p
부
@ARTICLE{e106-d_2_252,
author={Naoya MURAMATSU, Hai-Tao YU, Tetsuji SATOH, },
journal={IEICE TRANSACTIONS on Information},
title={Combining Spiking Neural Networks with Artificial Neural Networks for Enhanced Image Classification},
year={2023},
volume={E106-D},
number={2},
pages={252-261},
abstract={With the continued innovation of deep neural networks, spiking neural networks (SNNs) that more closely resemble biological brain synapses have attracted attention because of their low power consumption. Unlike artificial neural networks (ANNs), for continuous data values, they must employ an encoding process to convert the values to spike trains, suppressing the SNN's performance. To avoid this degradation, the incoming analog signal must be regulated prior to the encoding process, which is also realized in living things eg, the basement membranes of humans mechanically perform the Fourier transform. To this end, we combine an ANN and an SNN to build ANN-to-SNN hybrid neural networks (HNNs) that improve the concerned performance. To qualify this performance and robustness, MNIST and CIFAR-10 image datasets are used for various classification tasks in which the training and encoding methods changes. In addition, we present simultaneous and separate training methods for the artificial and spiking layers, considering the encoding methods of each. We find that increasing the number of artificial layers at the expense of spiking layers improves the HNN performance. For straightforward datasets such as MNIST, similar performances as ANN's are achieved by using duplicate coding and separate learning. However, for more complex tasks, the use of Gaussian coding and simultaneous learning is found to improve the accuracy of the HNN while lower power consumption.},
keywords={},
doi={10.1587/transinf.2021EDP7237},
ISSN={1745-1361},
month={February},}
부
TY - JOUR
TI - Combining Spiking Neural Networks with Artificial Neural Networks for Enhanced Image Classification
T2 - IEICE TRANSACTIONS on Information
SP - 252
EP - 261
AU - Naoya MURAMATSU
AU - Hai-Tao YU
AU - Tetsuji SATOH
PY - 2023
DO - 10.1587/transinf.2021EDP7237
JO - IEICE TRANSACTIONS on Information
SN - 1745-1361
VL - E106-D
IS - 2
JA - IEICE TRANSACTIONS on Information
Y1 - February 2023
AB - With the continued innovation of deep neural networks, spiking neural networks (SNNs) that more closely resemble biological brain synapses have attracted attention because of their low power consumption. Unlike artificial neural networks (ANNs), for continuous data values, they must employ an encoding process to convert the values to spike trains, suppressing the SNN's performance. To avoid this degradation, the incoming analog signal must be regulated prior to the encoding process, which is also realized in living things eg, the basement membranes of humans mechanically perform the Fourier transform. To this end, we combine an ANN and an SNN to build ANN-to-SNN hybrid neural networks (HNNs) that improve the concerned performance. To qualify this performance and robustness, MNIST and CIFAR-10 image datasets are used for various classification tasks in which the training and encoding methods changes. In addition, we present simultaneous and separate training methods for the artificial and spiking layers, considering the encoding methods of each. We find that increasing the number of artificial layers at the expense of spiking layers improves the HNN performance. For straightforward datasets such as MNIST, similar performances as ANN's are achieved by using duplicate coding and separate learning. However, for more complex tasks, the use of Gaussian coding and simultaneous learning is found to improve the accuracy of the HNN while lower power consumption.
ER -