|
Overview:
With the advance of information age, limited network capacity is under tremendous pressure due to the proliferation of information and the rapid development of hardware processing capability. How to reconcile the contradictions? In-network computation (or distributed computation of functions of distributed data) may give us an answer. Computation appears through the communication network, from application layer, network layer, to physical layer and it is easy to implement due to the advanced electronic technology. The basic idea of In-network Computation is to distribute the work of computation on each node according to the desired function, so that to reduce the flow of the network.
As a scalable and robust algorithm, gossip algorithm is an effective method to implement in-network computation. It has been proved that gossip algorithm is order-optimal in terms of computation rate. However, there are still many problems when it is applied in specific scenarios. One of our current research topics is to solve how to use gossip algorithms to compute in heterogeneous networks.
People: Jun Wang
Papers:
Jun Wang and Xudong Wang, "Gossip algorithm in heterogeneous networks", in preparation.
|