NUTS-BSNN: A non-uniform time-step binarized spiking neural network with energy-efficient in-memory computing macro

Dinh，Van Ngoc1; Bui，Ngoc My1; Nguyen，Van Tinh2; John，Deepu3; Lin，Long Yang4; Trinh，Quang Kien5

2023-12-01

10.1016/j.neucom.2023.126838

Neurocomputing   影响因子和分区

0925-2312

1872-8286

This work introduces a network architecture NUTS-BSNN: A Non-uniform Time-step Binarized Spiking Neural Network. NUTS-BSNN is a fully binarized spiking neural network with all binary weights, including the input and output layers. In the input and output layers, the weights are represented as stochastic series of numbers, while in the hidden layers, they are approximated to binary values for using simple XNOR-based computations. To compensate for the information loss due to binarization, we increased the convolutions at the input layer sequentially computed over multiple time-steps. The results from these operations are accumulated before generating spikes for the subsequent layers to increase the overall performance. We chose 14 time-steps for accumulation to achieve a good tradeoff between performance and inference latency. The proposed technique was evaluated using three datasets by direct training method and using a surrogate gradient algorithm. We achieved classification accuracies of 93.25%, 88.71%, and 70.31% on the Fashion-MNIST, CIFAR-10, and CIFAR-100 datasets, respectively. Further, we present an in-memory computing architecture for NUTS-BSNN, which limits resource and power consumption for hardware implementation.

Binary Spiking Neural Networks Edge-AI Applications In-memory Computing Neuromorphic Computing

EI ; SCI

英语

其他

20234114850516

Classification (of information) ; Energy efficiency ; Memory architecture ; Multilayer neural networks ; Stochastic systems

Energy Conservation:525.2 ; Information Theory and Signal Processing:716.1 ; Computer Systems and Equipment:722 ; Control Systems:731.1 ; Information Sources and Analysis:903.1 ; Systems Science:961

COMPUTER SCIENCE

2-s2.0-85173565269

Scopus

1.Academy of Military Science and Technology,Hanoi,Viet Nam
2.Nara Institute of Science and Technology,Japan
3.University College Dublin,Ireland
4.School of Microelectronics,Southern University of Science and Technology,Shenzhen,China
5.Faculty of Radio-Electronics – Le Quy Don Technical University,Hanoi,Viet Nam

Dinh，Van Ngoc,Bui，Ngoc My,Nguyen，Van Tinh,et al. NUTS-BSNN: A non-uniform time-step binarized spiking neural network with energy-efficient in-memory computing macro[J]. Neurocomputing,2023,560.

Dinh，Van Ngoc,Bui，Ngoc My,Nguyen，Van Tinh,John，Deepu,Lin，Long Yang,&Trinh，Quang Kien.(2023).NUTS-BSNN: A non-uniform time-step binarized spiking neural network with energy-efficient in-memory computing macro.Neurocomputing,560.

Dinh，Van Ngoc,et al."NUTS-BSNN: A non-uniform time-step binarized spiking neural network with energy-efficient in-memory computing macro".Neurocomputing 560(2023).