With the rapid development of edge computing, machine learning and deep learning approaches have been emerged to overcome the shortcomings of inappropriate offloading decisions. Edge server in the proximity of mobile users performs computation offloading, and it makes offloading decisions based on task information for each device and reduces the latency and energy consumption. Existing machine learning and deep learning works designed efficient learning algorithms for making offloading task decisions, but resource-constrained edge devices cannot effectively meet the requirements of Internet of Things (IoT) applications and Deep Neural Network (DNN) computing. Thus, Deep Reinforcement Learning (DRL)-based incentive mechanism in edge learning effectively makes the offloading tasks without knowing any prior information of the Edge Nodes under the dynamic network and improves the efficiency of task processing.