Control Techniques for Complex Networks教程.PDFVIP

Control Techniques for Complex Networks Sean Meyn LIDS Special Senimar Monday’s from 5-6pm (all semester- in Bd32- Room 144) Introduction and outline The stochastic networks community has become increasingly diverse and creative over the past decade. A biased sample of examples includes the choice of simplicity over optimality in the MaxWeight policy, relaxation techniques emerging from the theory of networks in heavy traffic, the introduction of concepts from economics, and the merging of simulation and stability techniques for performance evaluation. In these lectures we will explore many of these issues. The material is based on a book of the same name to appear in Cambridge University Press, 2007. Sample chapters will be made available to those with interest. The following three topics are personal favorites, but the plan can change depending on the preferences of the audience. (i) Why does the MaxWeight policy work? How can it be generalized? How can it be evaluated? We will consider broad generalizations, and touch on asymptotic optimality based on a relaxation technique introduced in the PhD thesis of Neil Laws, as well as techniques based upon a ”fluid model”. (ii) Can I understand the structure of an optimal policy for a complex network? Beau- tiful structure is revealed through a ”workload relaxation”, as well as truly practical approaches to design. (iii) Stochastic Lyapunov theory is the foundation for much of this research. This theory also has tremendous value for improving simulation algorithms. It is conjectured that the same techniques can be used to speed machine learning algorithms. For sample chapters and other references please contact me by email. The book and lectur