Strongly log concave polynomials, high dimensional simplicial complexes, and an FPRAS for counting Bases of Matroids

Shayan Oveis Gharan
University of Washington
February 25, 2019

A matroid is an abstract combinatorial object which generalizes the notions of spanning trees,
and linearly independent sets of vectors. I will talk about an efficient algorithm based on the Markov Chain Monte Carlo technique
to approximately count the number of bases of any given matroid. 

The proof is based on a new connections between high dimensional simplicial complexes, and a new class
of multivariate polynomials called completely log-concave polynomials. In particular, we exploit a fundamental fact from our

Positive geometries

Thomas Lam
University of Michigan; von Neumann Fellow, School of Mathematics
February 25, 2019

Positive geometries are real semialgebraic sets inside complex varieties characterized by the existence of a meromorphic top-form called the canonical form. The defining property of positive geometries and their canonical forms is that the residue structure of the canonical form matches the boundary structure of the positive geometry. A key example of a positive geometry is a projective polytope. 

Higher symplectic capacities

Kyler Siegel
Columbia University
February 25, 2019
I will describe a new family of symplectic capacities defined using rational symplectic field theory.
These capacities are defined in every dimension and give state of the art obstructions for various "stabilized" symplectic embedding problems such as one ellipsoid into another. They can also be described via symplectic cohomology and are related to counting pseudoholomorphic curves with tangency conditions. I will explain the basic idea of the construction and then give some computations, structural results, and applications.

Troubling Trends in ML Scholarship

Zachary Lipton
Carnegie Mellon University
February 22, 2019
Deep learning has led to rapid progress in open problems of artificial intelligence—recognizing images, playing Go, driving cars, automating translation between languages—and has triggered a new gold rush in the tech sector. But some scientists raise worries about slippage in scientific practices and rigor, likening the process to “alchemy.” How accurate is this perception? And what should the field do to combine rapid innovation with solid science and engineering?

The Epistemology of Deep Learning

Yann LeCun
Facebook, New York University
February 22, 2019
Deep learning has led to rapid progress in open problems of artificial intelligence—recognizing images, playing Go, driving cars, automating translation between languages—and has triggered a new gold rush in the tech sector. But some scientists raise worries about slippage in scientific practices and rigor, likening the process to “alchemy.” How accurate is this perception? And what should the field do to combine rapid innovation with solid science and engineering?

Reproducible, Reusable, and Robust Reinforcement Learning

Joelle Pineau
Facebook, McGill University
February 22, 2019
Deep learning has led to rapid progress in open problems of artificial intelligence—recognizing images, playing Go, driving cars, automating translation between languages—and has triggered a new gold rush in the tech sector. But some scientists raise worries about slippage in scientific practices and rigor, likening the process to “alchemy.” How accurate is this perception? And what should the field do to combine rapid innovation with solid science and engineering?

Institute Welcome

Robbert Dijkgraaf
IAS
February 22, 2019
Deep learning has led to rapid progress in open problems of artificial intelligence—recognizing images, playing Go, driving cars, automating translation between languages—and has triggered a new gold rush in the tech sector. But some scientists raise worries about slippage in scientific practices and rigor, likening the process to “alchemy.” How accurate is this perception? And what should the field do to combine rapid innovation with solid science and engineering?

Surrogates

Shai Shalev-Shwartz
Hebrew University of Jerusalem
February 22, 2019

Deep learning has led to rapid progress in open problems of artificial intelligence—recognizing images, playing Go, driving cars, automating translation between languages—and has triggered a new gold rush in the tech sector. But some scientists raise worries about slippage in scientific practices and rigor, likening the process to “alchemy.” How accurate is this perception? And what should the field do to combine rapid innovation with solid science and engineering?

Successes and Challenges in Neural Models for Speech and Language

Michael Collins
Google Research, Columbia University
February 22, 2019
Deep learning has led to rapid progress in open problems of artificial intelligence—recognizing images, playing Go, driving cars, automating translation between languages—and has triggered a new gold rush in the tech sector. But some scientists raise worries about slippage in scientific practices and rigor, likening the process to “alchemy.” How accurate is this perception? And what should the field do to combine rapid innovation with solid science and engineering?

Public programming: Panel discussion

Various Speakers
February 22, 2019
Deep learning has led to rapid progress in open problems of artificial intelligence—recognizing images, playing Go, driving cars, automating translation between languages—and has triggered a new gold rush in the tech sector. But some scientists raise worries about slippage in scientific practices and rigor, likening the process to “alchemy.” How accurate is this perception? And what should the field do to combine rapid innovation with solid science and engineering?