Stochastic Geometry to Generalize the Mondrian Process

Explore the intersection of stochastic geometry and machine learning in this 28-minute conference talk by Ngoc Mai Tran from the Hausdorff Center for Mathematics. Delve into the Mondrian process, a stochastic process that creates recursive partitions of space using random axis-aligned cuts. Discover how random forests and Laplace kernel approximations derived from the Mondrian process have led to efficient online learning methods and Bayesian optimization. Learn how viewing the Mondrian process as a special case of the stable under iterated tessellation (STIT) process allows for the application of stochastic geometry tools to address fundamental questions in machine learning. Gain insights into the main results and key questions arising from this novel intersection of fields. The talk covers topics such as introduction, Mondrian process, machine learning applications, stick trees, formal statements, and kernel methods, presenting joint work with Eliza O'Reilly.