DIKU Bits: Stochastic convex optimization

Speaker

Amir Yehudayoff, Professor in the Algorithms and Complexity (AC) section

Title

Stochastic convex optimization

Abstract

How should we mathematically define “learning”? There are several standard definitions, and each has its pros and cons. Stochastic convex optimization (SCO) provides one such definition. It is a benchmark framework for studying machine learning that is widely used for investigating stochastic optimization algorithms. We shall provide a brief introduction to SCO, and mostly discuss the role empirical risk minimization (ERM) plays in it.

Which courses do you teach?

I taught Advanced Topics in Machine Learning, and will be teaching Computability and Complexity, and also Approximation Algorithms.

Which technology/research/projects/startup are you excited to see the evolution of?

The company I am excited to see the evolution of is HT BioImaging.

What is your favorite sketch from the DIKUrevy?

The DIKUrevy I like most is Cafeen because I like coffee.