UW – MSR Machine Learning workshop 2015 – Session 5

15:30Towards Usable Machine Learning – Saleema Amershi People are integrally involved in the practice of machine learning, from providing machine learning algorithms with relevant data to examining and debugging machine learning performance. As a result, the ability of people to effectively train machine learning algorithms directly impacts the performance of the models being built. While advances continue in improving the performance of machine learning algorithms in the face of human and data limitations, a complementary strategy is needed to improve the capabilities of the people who are fundamentally involved in building and using machine learning based systems. In this talk, I will give examples of tools we are developing in the CHIL (Computer Human Interactive Learning) group at Microsoft Research to support people in the practice of machine learning and discuss some of the open challenges and opportunities in working towards more usable machine learning.

15:55Constructive Discrepancy Minimization for Convex Sets – Thomas Rothvoss A classical theorem of Spencer shows that any set system with n sets and n elements admits a coloring of discrepancy O(n1/2). Recent exciting work of Bansal, Lovett and Meka shows that such colorings can be found in polynomial time. In fact, the Lovett-Meka algorithm finds a half integral point in any “large enough” polytope. However, their algorithm crucially relies on the facet structure and does not apply to general convex sets.

16:20Spotlight: Submodular Point Processes – Rishabh Iyer We show that for any symmetric convex set K with measure at least exp(-n/500), the following algorithm finds a point y in K cap [-1,1]n with Omega(n) coordinates in -1,+1: (1) take a random Gaussian vector x; (2) compute the point y in K cap [-1,1]n that is closest to x. (3) return y.

16:25Closing – Ran Gilad-Bachrach This provides another truly constructive proof of Spencer’s theorem and the first constructive proof of a Theorem of Gluskin and Giannopoulos.

Date:
Speakers:
Thomas Rothvoss, Sebastien Bubeck, Rishabh K Iyer, and Ran Gilad-Bachrach
Affiliation:
University of Washington, Microsoft Research