Dynamic Resource Allocation Based On Deep Learning In Wireless Network Virtualization

With the rapid development of Internet and application communication,wireless network virtualization has become an important concept in the future network.It can provide multiple virtual wireless networks for different mobile virtual network operators(MVNOs)on the same physical infrastructure.Resource allocation is a major challenge for wireless virtualization.MVNO is more interested in maximizing its revenue.Because the price is unknown before calculation,the revenue maximization auction problem is much more complex.In wireless virtual network resource allocation,SP(service provider)sends a request to MVNO(mobile virtual network operator),MVNO sends a lease request to In P(infrastructure leasing provider)according to the request from SP,so as to obtain the corresponding network resources,and finally allocates the network resources obtained from the lease request to sp.In view of the above problems,MVNO resources need to be allocated most effectively.The main contents of this paper are as follows.First of all,in order to maximize social welfare,we need to allocate the resources in the network reasonably and dynamically.A resource allocation mechanism based on bilateral auction is proposed to meet the requirement of inter chip isolation in wireless networks.By dividing the buyer's SP into different types and designing different expressions with different priorities,the subsequent auction can be carried out smoothly after transformation,and finally the optimal resource allocation scheme is obtained.Secondly,operators pay more attention to their income,so to maximize the income,deep learning is used to design the best auction for MVNO resource allocation,and neural network architecture is built to provide accurate fitting for the best auction.Firstly,the neural network performs the monotonic transformation of valuation,and then calculates the allocation rules,i.e.the winning probability and conditional payment rules.Finally,the neural network is trained to adjust the parameters of the neural network to optimize the loss function,so as to optimize the MVNO resource allocation.Finally,in view of this paper have put forward based on bilateral auction allocation of resources and resource allocation based on the deep learning two strategies,has carried on the corresponding simulation experiments,used in evaluating strategies,and,in the wireless network virtualization environment carried out extensive simulation studies,to demonstrate the effectiveness of the proposed scheme,and comparing the experimental data,the validation and analysis,Finally,the effectiveness of the scheme is verified.