Abstract
Edge computing brings artificial intelligence algorithms and graphics processing units closer to data sources, making autonomy and energy-efficient processing vital for their design. Approximate computing has emerged as a popular strategy for energy-efficient circuit design, where the challenge is to achieve the best tradeoff between design efficiency and accuracy. The essential operation in artificial intelligence algorithms is the general matrix multiplication (GEMM) operation comprised of matrix multiplication and accumulation. This paper presents an approximate general matrix multiplication (AGEMM) unit that employs approximate multipliers to perform matrix–matrix operations on four-by-four matrices given in sixteen-bit signed fixed-point format. The synthesis of the proposed AGEMM unit to the 45 nm Nangate Open Cell Library revealed that it consumed only up to 36% of the area and 25% of the energy required by the exact general matrix multiplication unit. The AGEMM unit is ideally suited to convolutional neural networks, which can adapt to the error induced in the computation. We evaluated the AGEMM units’ usability for honeybee detection with the YOLOv4-tiny convolutional neural network. The results implied that we can deploy the AGEMM units in convolutional neural networks without noticeable performance degradation. Moreover, the AGEMM unit’s employment can lead to more area- and energy-efficient convolutional neural network processing, which in turn could prolong sensors’ and edge nodes’ autonomy.
Highlights
Artificial-intelligence-powered edge computing has brought complex processing devices closer to the data source, compromising their autonomy [1]
Even though the unit’s core design was equal for all multipliers, we differentiated between the general matrix multiplication (GEMM) unit with the exact multiplier and the approximate general matrix multiplication (AGEMM) unit with an approximate multiplier for clarity
High values of the mAP[0.5] metric indicate that the object detector performs well, while lower values of the mAP[0.5:0.95] metric suggest that the detector is not very good at localization
Summary
Artificial-intelligence-powered edge computing has brought complex processing devices closer to the data source, compromising their autonomy [1]. The most recent study [11] was the first to use deep neural network object detectors implemented on graphics processing units for Varroa destructor mite detection on a honeybee. All these solutions were based on offline processing of the recorded images or videos and lacked permanent monitoring performed near beehives, commonly without a power supply, ensured only by a long-term autonomy device. The two-stage detectors include various correlated phases such as region proposal generation, feature extraction using convolutional neural networks, bounding box regression, and classification, which are trained separately. Used single-stage detectors are the you look only once detector (YOLO) [73], the Single-shot multi-box detector (SSD) [74], and RetinaNET [75]
Sign-in/Register to view highlights & summary 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.
