Echo state network (ESN) as a novel artificial neural network has drawn much attention from time series prediction in edge intelligence. ESN is slightly insufficient in long-term memory, thereby impacting the prediction performance. It suffers from a higher computational overhead when deploying on edge devices. We firstly introduce the knowledge distillation into the reservoir structure optimization, and then propose the echo state network based on improved knowledge distillation (ESN-IKD) for edge intelligence to improve the prediction performance and reduce the computational overhead. The model of ESN-IKD is constructed with the classic ESN as a student network, the long and short-term memory network as a teacher network, and the ESN with double loop reservoir structure as an assistant network. The student network learns the long-term memory capability of the teacher network with the help of the assistant network. The training algorithm of ESN-IKD is proposed to correct the learning direction through the assistant network and eliminate the redundant knowledge through the iterative pruning. It can solve the problems of error learning and redundant learning in the traditional knowledge distillation process. Extensive experimental simulation shows that ESN-IKD has a good time series prediction performance in both long-term and short-term memory, and achieves a lower computational overhead.

In the 5G environment, the edge computing paradigm enables service providers to deploy their service instances on distributed edge servers to serve nearby end users with extremely low latency. This boosts the emergence of modern applications, like AR/VR, online gaming, and autonomous vehicles. Existing approaches find service provision strategies under the assumption that all the user requirements are known. However, this assumption may not be true in practice and thus the effectiveness of existing approaches could be undermined. Inspired by the great success of recommender systems in various fields, we can mine users’ interests in new services based on their similarities in terms of current service usage. Then, new service instances can be provisioned accordingly to better fulfil users’ requirements. We formulate the problem studied in this paper as a Cost-aware Recommendation-oriented Edge Service Provision (CRESP) problem. Then, we formally model the CRESP problem as a Constrained Optimization Problem (COP). Next, we propose CRESP-O to find optimal solutions to small-scale CRESP problems. Besides, to solve large-scale CRESP problems efficiently, we propose an approximation approach named CRESP-A, which has a theoretical performance guarantee. Finally, we experimentally evaluate the performance of both CRESP-O and CRESP-A against several state-of-the-art approaches on a public testbed.