C-DNN: A 24.5-85.8TOPS/W Complementary-Deep-NeuralNetwork Processor with Heterogeneous CNN/SNN Core Architecture and Forward-Gradient-Based Sparsity Generation

📋 论文概要

提出一种异构CNN/SNN核心架构的深度神经网络处理器C-DNN，通过互补利用CNN和SNN的优势，实现24.5-85.8 TOPS/W的高能效。解决了SNN能量随层间尖峰稀疏性波动的问题，同时保持高准确率。

💡 主要创新点

• 首次提出互补深度神经网络（C-DNN）处理器，在同一芯片上集成CNN和SNN异构核心，根据层间尖峰稀疏性动态选择计算模式。
• 通过事件驱动与帧驱动混合计算，优化能量效率，实现24.5-85.8 TOPS/W的峰值能效。

🏷 关键词

📄 原文摘要

have been shown to achieve the same accuracy as Convolutional-Neural-Networks (CNNs). By using CNN-to-SNN conversion, SNNs become a promising candidate for ultra-low power AI applications [1]. For example, compared to BNNs or XOR-nets, SNNs provide lower power consumption and higher accuracy [2]. This is because SNNs perform spikebased event-driven operation with high spike sparsity, unlike a CNN’s frame-driven operation. Fig. 22.5.1 shows that the energy consumption of a SNN fluctuates up and down along the layers depending on spike sparsity which changes with each layer, whereas a CNN shows comparatively lower variation. Also, SNNs offer low-power training by generating a Forward-Gradient (FG) which is computed as the time difference between a pre-spike and post-spike similar to STDP in a biological neuron [3]. However, SNN

👥 作者与机构

Sangyeob Kim, Soyeon Kim, Seongyon Hong, Sangjin Kim, Donghyeon Han, Hoi-Jun Yoo

Korea Advanced Institute of Science and Technology, Daejeon, Korea Spiking-Neural-Networks (SNNs) have been studied for a long time, and recently