arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2009.07190cs.CV
收藏

低硬件要求的ResNet-like体系结构

ResNet-like Architecture with Low Hardware Requirements

FRC CSC RAS(俄罗斯科学院计算机系统与控制研究所(FRC CSC RAS));Smart Engines Service LLC(Smart Engines服务有限责任公司);JSC MCST(MCST股份公司);Institute for Information Transmission Problems RAS(俄罗斯科学院信息传输问题研究所)

Elena Limonova, Daniil Alfonso, Dmitry Nikolaev, Vladimir V. Arlazarov

摘要:

现代识别系统中计算最密集的部分之一是用于图像分类、分割、增强和识别的深度神经网络的推理。边缘计算的日益普及促使我们寻找方法来减少移动和嵌入式设备的时间。减少神经网络推理时间的一种方法是修改神经元模型,使其在特定设备上的计算更有效。这种模型的例子是双极形态神经元模型。双极形态神经元基于用加法和最大值运算代替乘法的思想。该模型已被证明用于LeNet类架构的简单图像分类[1]。在本文中,我们介绍了一个双极形态ResNet(BM-ResNet)模型从一个更复杂的ResNet架构,通过转换其层到双极形态的。我们将BM-ResNet应用于MNIST和CIFAR-10数据集的图像分类,准确率仅适度下降,从99.3%降至99.1%,从85.3%降至85.1%。我们还估计了所得到的模型的计算复杂性。我们表明,对于大多数ResNet层,所考虑的模型需要2.1-2.9倍的逻辑门来实现,延迟降低15-30%。

英文摘要:

One of the most computationally intensive parts in modern recognition systems is an inference of deep neural networks that are used for image classification, segmentation, enhancement, and recognition. The growing popularity of edge computing makes us look for ways to reduce its time for mobile and embedded devices. One way to decrease the neural network inference time is to modify a neuron model to make it moreefficient for computations on a specific device. The example ofsuch a model is a bipolar morphological neuron model. The bipolar morphological neuron is based on the idea of replacing multiplication with addition and maximum operations. This model has been demonstrated for simple image classification with LeNet-like architectures [1]. In the paper, we introduce a bipolar morphological ResNet (BM-ResNet) model obtained from a much more complex ResNet architecture by converting its layers to bipolar morphological ones. We apply BM-ResNet to image classification on MNIST and CIFAR-10 datasets with only a moderate accuracy decrease from 99.3% to 99.1% and from 85.3% to 85.1%. We also estimate the computational complexity of the resulting model. We show that for the majority of ResNet layers, the considered model requires 2.1-2.9 times fewer logic gates for implementation and 15-30% lower latency.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑