Bayesian Classification With Gaussian Processes
IEEE Transactions on Pattern Analysis and Machine Intelligence
A Bayesian approach to on-line learning
On-line learning in neural networks
Approximation algorithms for directed Steiner problems
Proceedings of the ninth annual ACM-SIAM symposium on Discrete algorithms
A new approximation algorithm for the Steiner tree problem with performance ratio 5/3
Journal of Algorithms
Grid Coverage for Surveillance and Target Location in Distributed Sensor Networks
IEEE Transactions on Computers
A family of algorithms for approximate bayesian inference
A family of algorithms for approximate bayesian inference
Near-optimal sensor placements: maximizing information while minimizing communication cost
Proceedings of the 5th international conference on Information processing in sensor networks
Assessing Approximate Inference for Binary Gaussian Process Classification
The Journal of Machine Learning Research
| Hi-index | 0.00 |
We consider the problem of deploying relay nodes to achieve connectivity with minimum cost in a sensor network of unknown radio propagation characteristics. For a network where a certain number of targets or sensing nodes have already been deployed in fixed and known positions, we aim at efficiently adding communication or relay nodes to guarantee connectivity with minimum cost, between any sensor node and a base station. The communication cost of a wireless link is defined as the expected number of retransmissions over that link and is modeled using an underlying Gaussian Process (GP) between the nodes. We propose an iterative sensor deployment approach that learns the parameters of the underlying GP while deploying the additional nodes in the best positions possible at each step. Our deployment algorithm is more powerful with respect to the ones found in literature since: 1) we do not assume fixed communication range, i.e., we do not assume that nodes can perfectly communicate within a fixed range and will not communicate at all outside that range (this assumption is not realistic for the wireless channel); 2) we do not assume the existence of a pilot deployment aimed at learning the radio propagation characteristics because of the high cost of the deployment process and of the sensor nodes themselves.