Skip to yearly menu bar Skip to main content


Contributed talks
in
Workshop: Adaptive and Scalable Nonparametric Methods in Machine Learning

Debarghya Ghoshdastidar, Ulrike von Luxburg. Do Nonparametric Two-sample Tests work for Small Sample Size? A Study on Random Graphs.

[ ] [ Project Page ]
2016 Contributed talks

Abstract:

We consider the problem of two-sample hypothesis testing for inhomogeneous unweighted random graphs, where one has access to only a very small number of samples from each model. Standard tests cannot be guaranteed to perform well in this setting due to the small sample size. We present a nonparametric test based on comparison of the adjacency matrices of the graphs, and prove that the test is consistent for increasing sample size as well as when the graph size increases with sample size held fixed. Numerical simulations exhibit the practical significance of the test.

Live content is unavailable. Log in and register to view live content