Segmentation point data processing method of cloud edge collaborative neural network

• Main Authors: WANG WEI, GAO YU, WANG HUIQIONG, ZHANG CHAOYANG, QIU LANXIN
• Format: Patent
• Language: Chinese; English
• Published: 19.12.2023
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; ELECTRIC DIGITAL DATA PROCESSING; PHYSICS

The invention discloses a segmentation point data processing method for a cloud-edge collaborative neural network, and the method comprises the steps: adding a network exit module and a data compression module at a cloud-edge segmentation point of a neural network; the data at the segmentation points are selected to exit the neural network in advance, quantified according to a certain proportion and then uploaded to the cloud or the original data are uploaded to the cloud and other classification processing modes, a classification threshold iterative update algorithm is designed, and a classification threshold is obtained by combining the network communication condition and the data information entropy; and the inference accuracy of the neural network is maximized under a given cloud-side communication data volume. According to the method, the real-time performance and the accuracy of the hierarchical network under the condition that communication is limited are effectively improved, and the method can be use