Seminars Sorted by Series

Implicit generative models such as GANs have achieved remarkable progress at generating convincing fake images, but how well do they really match the distribution? Log-likelihood has been used extensively to evaluate generative models whenever it’s...

While the trend in machine learning has tended towards more complex hypothesis spaces, it is not clear that this extra complexity is always necessary or helpful for many domains. In particular, models and their predictions are often made easier to...

In this talk, I will first introduce our recent work on the Deep Equilibrium Model (DEQ). Instead of stacking nonlinear layers, as is common in deep learning, this approach finds the equilibrium point of the repeated iteration of a single non-linear...

Despite the success of deep learning, much of its success has existed in settings where the goal is to learn one, single-purpose function from data. However, in many contexts, we hope to optimize neural networks for multiple, distinct tasks (i.e...

Physical processes in the world often have a modular structure, with complexity emerging through combinations of simpler subsystems. Machine learning seeks to uncover and use regularities in the physical world. Although these regularities manifest...

Standard machine learning produces models that are highly accurate on average but that degrade dramatically when the test distribution deviates from the training distribution. While one can train robust models, this often comes at the expense of...

There are several broad insights we can draw from computational models of human cognition in order to build more human-like forms of machine learning. (1) The brain has a great deal of built-in structure, yet still tremendous need and potential for...

We are interested in moments of velocity increments and derivatives, characterized by scaling exponents overline{(v(x + r) − v(x))n} ∝ r^ζn and overline{(∂xvx)n} ∝ Re^ρn , respectively. In high Reynolds number flows, the moments of different orders...

Abstract: The confounding question of asymptotically high Rayleigh number heat transport in Rayleigh-Bénard convection modeled by the Boussineq approximation to the Navier-Stokes equations is reviewed from viewpoints of theory (models of the model)...

Abstract: I will discuss some preliminary work on using machine learning
to produce turbulence models that can be used in large eddy simulation.
I will discuss how better models can be constructed and in general,
how one can use machine learning to...

5 minutes each:Steve Childress (NYU)Camillo de Lellis (IAS)Yakov Sinai (Princeton University/IAS)Jalal Shatah (NYU)Larry Sirovich (Rockefeller)John Wettlaufer (Yale University)

Abstract: I will discuss the recent non-uniqueness result with Vlad Vicol on the non-uniqueness of weak solutions to the Navier-Stokes equations, as well as the follow up paper by myself, Maria Colombo and Vicol. I hope to phrase the results within...

Abstract: We consider a class of dynamical systems described by ordinary differential equations with an isolated singularity, where the singularity is characterized by the lack of Lipschitz continuity. Singularities are common in applications both...

Abstract: The purpose of this work is to perform a mathematically rigorous study of Lagrangian chaos and "passive scalar turbulence" in incompressible fluid mechanics. We study the Lagrangian flow map associated to 2D Navier-Stokes and hyper-viscous...

Abstract: I will present a new approach to the isometric embedding problem. The main new idea is to use the theory of stochastic flows in combination with various possible gradient flow structures. These ideas are motivated by the statistical...