When an epidemic spreads in a network, being able to detect its source, i.e., the node that initiated the propagation, is extremely useful. For example, this could help to develop appropriate containment measures or to prevent future epidemics. In this project, we will work on sensor based source localization; a model in which we observe the infection times of a few sensor nodes in the network.
In the past decade, a number of algorithms were developed for source localization, most of which are heuristic in nature. As it is in other areas of machine learning, it is sometimes difficult to reproduce the results published in each paper, yet alone to make a fair comparison between them. In the INDY lab, we developed a benchmarking system that allows this comparison, however, there are still a large number of algorithms that have not been added to the framework.
Over the course of the semester, the student will understand and implement several existing algorithms. There will be substantial guidance by the supervising PhD student in selecting and understanding these algorithms. By the end of the semester the student will have a good understanding about the state of the art methods in the field of source localization.
The student will learn to use our benchmarking system; a useful experience at a time when reproducibility in machine learning is becoming more and more important.