Student: Hong Li, PhD student
Supervisors: Prof. Lorne Mason & Prof. Michael Rabbat
Abstract: We developed active probing strategies and applied stochastic automata to learn the best paths from
feedback of the network. The simulation results show that the active probing and learning method can find a better path than the minimum hop path in
terms of the mean network delay. It takes around 1000-2000 probes for the learning automata with an appropriate gain to learn the optimal paths. The
convergence speed scales well with overlay network sizes.
[Full Description] [Paper (pdf format)]
|