边缘装置的多元时序异常检测模型部署方法

Deployment of Multiple Timing Anomaly Detection Model for Edge Devices

为提升时间序列数据的处理质量和精度,神经网络技术被应用于相关领域。然而,神经网络推理需要大量的硬件计算,边缘设备难以满足要求。文章选用MTGNN网络作为研究目标,通过基于模块替换的网络压缩方法对目标网络进行压缩,在保持精度的基础之上缩减神经网络的计算量。

To improve the processing quality and accuracy of time series data,neural network technology has been applied in related fields.However,neural network reasoning requires a lot of hardware computing,and Edge device are difficult to meet the requirements.This study selects the MTGNN network as the research objective,and compresses the target network through a module replacement based network compression method,reducing the computational complexity of the neural network while maintaining accuracy.