Abstract
Applications of neural networks have gained significant importance in embedded mobile devices and Internet of Things (IoT) nodes. In particular, convolutional neural networks have emerged as one of the most powerful techniques in computer vision, speech recognition, and AI applications that can improve the mobile user experience. However, satisfying all power and performance requirements of such low power devices is a significant challenge. Recent work has shown that binarizing a neural network can significantly improve the memory requirements of mobile devices at the cost of minor loss in accuracy. This paper proposes MB-CNN, a memristive accelerator for binary convolutional neural networks that perform XNOR convolution in-situ novel 2R memristive data blocks to improve power, performance, and memory requirements of embedded mobile devices. The proposed accelerator achieves at least 13.26 × , 5.91 × , and 3.18 × improvements in the system energy efficiency (computed by energy × delay) over the state-of-the-art software, GPU, and PIM architectures, respectively. The solution architecture which integrates CPU, GPU and MB-CNN outperforms every other configuration in terms of system energy and execution time.
Highlights
Sensor equipped Internet of Things (IoT) devices are expected to impact the future of consumer electronics
All of the relevant parameters of a trained binary convolutional neural network (B-convolutional neural network (CNN)) model are first written into the Resistive RAM (RRAM) arrays, which can be later used for inference tasks many times
The true and complement forms of every filter element are stored in a cell as b and b, respectively, where logic 1 is represented by the low resistance state (LRS) of the RRAM cell, and the high resistance state (HRS) is used for 0
Summary
Sensor equipped Internet of Things (IoT) devices are expected to impact the future of consumer electronics. The increased demand of computation, memory, and energy consumption creates serious challenges to its applicability in IoT mobile devices. Recent studies have shown that the memory requirements of neural networks can be reduced by applying various compression and quantization techniques [7,8]. Binary neural network [9] and XNOR-Net [10] replace the power consuming floating point multiplications with bitwise XNOR operations. One bit quantization helps to achieve significant performance improvements for state-of-the-art neural networks. Because of its high density and low read latency, all the parameters of a huge neural network can be stored without impacting the overall area or performance of the device. The integration of MB-CNN accelerator with a single core CPU device provides 4.14× improvement in performance and 4× improvement in energy over software based single core CPU solution. As compared with GPU based and PIM-like accelerators, the proposed accelerator achieves at least 5.91× and 3.18× improvements in the system energy-delay-product, respectively
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
