A 64K-Neuron 64M-1b-Synapse 2.64pJ/SOP Neuromorphic Chip With All Memory on Chip for Spike-Based Models in 65nm CMOS

Kuang, Yisong; Cui, Xiaoxin; Zhong, Yi; Liu, Kefei; Zou, Chenglong; Dai, Zhenhui; Wang, Yuan; Yu, Dunshan; Huang, Ru

doi:10.1109/tcsii.2021.3052172

Cited by 40 publications

(22 citation statements)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This custom integrated circuit (IC) should support parallel computing on multiple levels to decrease the compute latency (Chen, Krishna, Emer and Sze (2016)). In addition, the low firing rate of the proposed SRNN model also needs to be fully used by the designed IC to reduce the calculation and energy consumption (Kuang, Cui, Zhong, Liu, Zou, Dai, Wang, Yu and Huang (2021)). A fully functional and powerful IC can solve the above problems, which is the core component of retinal prostheses.…”

Section: Discussionmentioning

confidence: 99%

SpikeSEE: An Energy-Efficient Dynamic Scenes Processing Framework for Retinal Prostheses

Fang¹,

Zou²,

Yang³

2022

Preprint

View full text Add to dashboard Cite

Intelligent and low-power retinal prostheses are highly demanded in this era, where wearable and implantable devices are used for numerous healthcare applications. In this paper, we propose an energy-efficient dynamic scenes processing framework (SpikeSEE) that combines a spike representation encoding technique and a bio-inspired spiking recurrent neural network (SRNN) model to achieve intelligent processing and extreme low-power computation for retinal prostheses. The spike representation encoding technique could interpret dynamic scenes with sparse spike trains, decreasing the data volume. The SRNN model, inspired by the human retina's special structure and spike processing method, is adopted to predict the response of ganglion cells to dynamic scenes. Experimental results show that the Pearson correlation coefficient of the proposed SRNN model achieves 0.93, which outperforms the state-of-the-art processing framework for retinal prostheses. Thanks to the spike representation and SRNN processing, the model can extract visual features in a multiplication-free fashion. The framework achieves 12 times power reduction compared with the convolutional recurrent neural network (CRNN) processing-based framework. Our proposed SpikeSEE predicts the response of ganglion cells more accurately with lower energy consumption, which alleviates the precision and power issues of retinal prostheses and provides a potential solution for wearable or implantable prostheses.

show abstract

Section: Discussionmentioning

confidence: 99%

SpikeSEE: An Energy-Efficient Dynamic Scenes Processing Framework for Retinal Prostheses

Fang¹,

Zou²,

Yang³

2022

Preprint

View full text Add to dashboard Cite

show abstract

“…Moreover, neuromorphic computing has reached some significant milestones, which have been well summarized in [ 3 ]. At the time of publication for this paper, a few neuromorphic processors have already been manufactured and well-utilized: TrueNorth [ 36 ], Loihi [ 37 ], SpiNNaker [ 38 ], and some others still in development [ 39 ]. These neuromorphic processors are based on a non-Von Neumann computer architecture and can be well suited with an event-based camera, because of the spiking, asynchronous output of events.…”

Section: Event-based Depth Estimation In Hardwarementioning

confidence: 99%

Analytical Review of Event-Based Camera Depth Estimation Methods and Systems

Furmonas

Liobe

Barzdėnas

2022

Sensors

View full text Add to dashboard Cite

Event-based cameras have increasingly become more commonplace in the commercial space as the performance of these cameras has also continued to increase to the degree where they can exponentially outperform their frame-based counterparts in many applications. However, instantiations of event-based cameras for depth estimation are sparse. After a short introduction detailing the salient differences and features of an event-based camera compared to that of a traditional, frame-based one, this work summarizes the published event-based methods and systems known to date. An analytical review of these methods and systems is performed, justifying the conclusions drawn. This work is concluded with insights and recommendations for further development in the field of event-based camera depth estimation.

show abstract

“…Each coding method has the optimal threshold and leakage value interval, and the network energy efficiency with better accuracy is close. (5) The input encoding and inter-layer data sparsity show a negative correlation with inference accuracy, a positive correlation with energy efficiency, and a more ambiguous correlation with adversarial robustness, where the most important influencing factor is in the distribution of weights. In order to achieve the same issuing condition with parameters that have lower input encoding and inter-layer data sparsity during training, larger weights of local neurons are needed.…”

Section: ) Summarymentioning

confidence: 99%

A Comparative Study on the Performance and Security Evaluation of Spiking Neural Networks

Cui

Zhou

et al. 2022

IEEE Access

Self Cite

View full text Add to dashboard Cite

The brain-inspired Spiking neural networks (SNN) claim to present advantages for visual classification tasks in terms of energy efficiency and inherent robustness. In this work, we explore the impact on network inter-layer sparsity through neural coding schemes and the intrinsic structural parameters of Leaky Integrate-and-Fire (LIF) neurons , which can be a candidate metric for performance evaluation. Towards this, we perform a comparative study of four critical neural coding schemes: rate coding (poisson coding), latency coding, phase coding, and direct coding, as well as 6 LIF neuron intrinsic parameter options for a total of 24 combined parameter schemes. Specifically, the models were trained using a supervised training algorithm with a surrogate gradient, and two adversarial attacks, Fast Gradient Sign Method (FGSM) and Projected Gradient Descent (PGD) were applied on a CIFAR10 dataset. We identified the sources of interlayer sparsity in SNN, and quantitatively analyzed the differences in sparsity caused by coding schemes, neuron leakage factors and thresholds. Various aspects of network performance were thoroughly considered in this paper, including inference accuracy, adversarial robustness, and energy efficiency. Our results show that latency coding is the optimum choice in achieving the highest adversarial robustness and energy efficient against low intensity attacks, while rate coding offers the best adversarial robustness against medium and high intensity attacks. The maximum deviations of robustness and efficiency between different coding schemes are 9.35% in VGG5 and 13.59% in VGG9. Increasing the sparsity of spike activity by improving the threshold can bring a short-lived adversarial robustness sweet spot, while excessive sparsity due to changes in threshold and leakage can instead reduce the adversarial robustness. The study reveals the advantages and disadvantages, and design space of SNN in various dimensions, allowing researchers to frame their neuromorphic systems in terms of the coding methods, neuron inherent structure, and model learning capabilities. INDEX TERMSSpiking neural network, Accuracy, Energy efficiency, Adversarial robustness, Sparsity I. INTRODUCTION S PIKING neural networks (SNN), which have spatiotemporal spiking sparsity and biological-like properties, are increasingly used to investigate more energy-efficient neural computing circuits than artificial neural networks (ANN) [1], [2]. Numerous studies have begun to use biologically plausible learning methods to implement neuromorphic circuits [3]-[5]. The SNN structure was discovered to have certain robust advantages in resisting sample noise and adversarial attacks due to its sparse inherent structure and discrete encoding of the input [6]-[9]. The main challenge in optimizing SNN is the lack of a way to achieve reliable metrics for their inference accuracy. Current performance improving methods can be broadly classified into three types: ANN-converted SNN [10], [11], surrogate gradient training methods [12], and unsupervised ...

show abstract

A 64K-Neuron 64M-1b-Synapse 2.64pJ/SOP Neuromorphic Chip With All Memory on Chip for Spike-Based Models in 65nm CMOS

Cited by 40 publications

References 25 publications

SpikeSEE: An Energy-Efficient Dynamic Scenes Processing Framework for Retinal Prostheses

SpikeSEE: An Energy-Efficient Dynamic Scenes Processing Framework for Retinal Prostheses

Analytical Review of Event-Based Camera Depth Estimation Methods and Systems

A Comparative Study on the Performance and Security Evaluation of Spiking Neural Networks

Contact Info

Product

Resources

About