Log-concave density estimation: adaptation and high dimensions

Richard Samworth
University of Cambridge

In recent years, density estimation via log-concave maximum likelihood estimation has emerged as a fascinating alternative to traditional nonparametric smoothing techniques, such as kernel density estimation, which require the choice of one or more bandwidths. I will outline some of the attractive properties of this technique, with a focus on new results on adaptivity properties and the estimation of high-dimensional log-concave densities.

Presentation (PDF File)

Back to Workshop III: Geometry of Big Data