Speaker: Ilias Diakonikolas (Edinburgh)

Title: Algorithmic aspects of inference

Abstract: We will focus on a core problem in statistics: how to estimate (learn) an unknown probability distribution based on random samples. Density estimation is one of the most fundamental inference problems, but much of the classic theory relies on estimators that are hard to compute in general.

In this talk, we will explore the following question: Are there alternatives for the maximum likelihood estimator that are statistically efficient, but can in addition be computed efficiently?

We will survey recent algorithmic ideas that lead to statistically and computationally efficient learning algorithms in both low- and high-dimensional settings.