Delay-sensitive tasks on intelligent vehicles require a high-level computing power.However,the computing resources available on the requested vehicles are not enough to process the entire task data alone,and can hardly meet the delay requirements.Vehicle fog computing(VFC)can improve vehicle service by offloading computation near the requesting vehicle.In this paper,the computation offloading process is modeled based on two-stage production planning,and a computation offloading algorithm(COA)is proposed to optimize the offloading scheme and execution sequence,thereby reducing the computation offloading delay.Based on the genetic algorithm(GA),COA applies Johnson Rules to determine the offloading order.Simulations on SUMO and MATLAB demonstrate that COA has lower average offloading latency and better stability than GA under the same number of iterations.
Small and medium-sized enterprises lack sufficient data and feature tags for user portraits in practice,which hinders them from obtaining accurate results.Plus,training the cross-industry and cross-departmental data in a centralized manner can lead to privacy leakage.In these regards,a multi-source data user portrait design scheme based on federated learning is proposed.The scheme designs the framework of the federated portrait system,and uses the computer mechanism of federated learning and the privacy intersection algorithm to realize multi-source data sharing.Thus,the privacy and security of the data sets of all parties are ensured.Experiments show that compared with the local portrait scheme and some other multi-source data portrait schemes,the proposed scheme can greatly improve the prediction accuracy,and has higher privacy security and algorithm scalability.
Session-based recommender systems(SRS)can predict a user's next click based on the historical behavior sequence of anonymous users.Most existing recommendation models ignore the potential relevance of the current session to other sessions and fail to capture the user's main purpose.To solve this problem,a session-based recommendation model(SR-BGEAN)is proposed based on Bi-GRU and external attention networks.Taking the external attention mechanism into the session recommendation task for the first time,the model combines Bi-GRU for modeling and captures both intra-session items bidirectional sequential information and inter-session sample global information to form the final interest representation.Experimental results on the Diginetica and Yoochoose datasets show that the SR-BGEAN model has a higher hit rate of 71.50％in the top 20 recommended items compared to the optimal baseline model.
For the indoor mixed power line communication(PLC)-visible light communication(VLC)/radio frequency(RF)transmission system under the decode-and-forward(DF)protocol,this paper proposes maximal ratio combining(MRC)and selection combining(SC)schemes,and analyzes the systems'outage performance.Firstly,the end-to-end signal-to-noise ratio(SNR)under MRC and SC schemes is presented by considering the impacts of the pass loss,the channel fading and the randomness of user's location.Secondly,the closed-form expressions of outage probability under these two schemes are derived.Finally,the computer simulation validates the theoretical formulas by numerical results,and analyzes the impacts of different parameters on the system outage performance.Our study provides a theoretical basis for the practical application of the considered mixed PLC-VLC/RF system.
To reduce the task completion time in mobile edge computing(MEC)systems,this paper proposes a computation offloading algorithm based on data compression and non-orthogonal multiple access(NOMA)technologies.The solution aims to minimize the task completion time by optimizing communication and computing resources.Thus,an algorithm based on the artificial fish swarm algorithm is proposed.Since the fixed step of the conventional artificial fish swarm algorithm leads to a slow convergence rate and a large amount of computation,and probably a local optimal solution,this paper proposes an adaptive step-size artificial fish swarm algorithm that can flexibly control the convergence rate.The simulation results show that the proposed algorithm can significantly reduce the task completion time compared to the conventional methods based on the data compression technology or the NOMA technology.
