论文标题
用于近似局部二进制模式网络的近传感器处理加速器
A Near-Sensor Processing Accelerator for Approximate Local Binary Pattern Networks
论文作者
论文摘要
在这项工作中，提出了基于高速和节能的近传感器局部二进制二进制二进制二进制二进制模式架构（NS-LBP）来执行一种新型的局部二进制二进制模式深神经网络。首先，灵感来自最近的LBP网络，我们设计了一个近似，面向硬件的和多重蓄积的（MAC）的无用网络，名为AP-LBP，用于有效的功能提取，进一步降低了计算复杂性。然后，我们开发NS-LBP作为SNSRAM中的处理单元和一种并行的内存LBP算法来处理缓存中传感器附近的图像，从而显着降低了数据传输的功率消耗到外芯片处理器。与基线CNN和LBP-NETWORK模型相比，我们对MNIST和SVHN数据集的电路对拟合结果的结果很小，而NS-LBP则达到1.25 GHz，能力效率为37.4 TOPS/W。与最近的基于LBP的最佳网络相比，NS-LBP将能耗降低2.2倍，执行时间为4倍。
In this work, a high-speed and energy-efficient comparator-based Near-Sensor Local Binary Pattern accelerator architecture (NS-LBP) is proposed to execute a novel local binary pattern deep neural network. First, inspired by recent LBP networks, we design an approximate, hardware-oriented, and multiply-accumulate (MAC)-free network named Ap-LBP for efficient feature extraction, further reducing the computation complexity. Then, we develop NS-LBP as a processing-in-SRAM unit and a parallel in-memory LBP algorithm to process images near the sensor in a cache, remarkably reducing the power consumption of data transmission to an off-chip processor. Our circuit-to-application co-simulation results on MNIST and SVHN data-sets demonstrate minor accuracy degradation compared to baseline CNN and LBP-network models, while NS-LBP achieves 1.25 GHz and energy-efficiency of 37.4 TOPS/W. NS-LBP reduces energy consumption by 2.2x and execution time by a factor of 4x compared to the best recent LBP-based networks.
