Machine Learning for Complex Networks

点击翻译按钮获取中文摘要

My proposed research program focuses on developing methods and tools for improving performance of complex networks, including data communication and social networks. The program encompasses analysis of data collected from deployed networks, characterization and modeling of network traffic, and development of software tools for evaluating network performance.My overall objective is to develop formal analytical and statistical methods for analysis of complex networks. I will employ spectral graph theory and machine learning to analyze various topologies of complex social networks such as Facebook, LinkedIn, Twitter, and their dynamical behavior. This analysis will capture historical trends in the development of these networks. Machine learning techniques and algorithms will be also used to classify network anomalies in complex data networks. Training and testing of large collected datasets will be performed using the Compute Canada Cedar national system located at Simon Fraser University.I also plan to apply machine learning techniques for developing new algorithms for resource allocations in software-defend networks (SDN) and network functions virtualization (NFV). The virtualized network architecture enables coexistence of multiple virtual networks on an existing physical infrastructure. The Virtual Network Embedding (VNE) problem, which deals with the embedding of virtual network components onto a physical network, is known to be non-deterministic polynomial-time (NP)-hard. Hence, of particular interest and challenge is the development of VNE algorithms for wireless networks where the network topology varies with time. Performance of the developed network algorithms will be evaluated using our recently developed VNE-Sim software platform that enables definition and implementation of various network elements and generation of various network topologies.Research results emanating from the proposed program will improve our understanding of the underlying mechanisms that govern the behavior of the Internet and social networks. They will enhance network security and help improve network performance. This is of fundamental importance to Canadian telecommunications industry and to Canadian network service providers. A library of developed software tools and models will be made publicly available to the research community.