Building Graph Structures from the Beta Process
Non-Parametric Bayes Workshop at NIPS 2009, 2009
We show how a variety of directed and undirected graphs can be modelled via the hierarchical Beta process. This extends the domain of non-parametric methods to an important new class of problems. We give experimental results for the task of inferring topological maps using data collected from robot sensors. Our results show that we can accurately infer a map from noisy data.