לוח שנה משולב

The Burger-Sarnak method shows that the restriction of an automorphic representation of a reductive group to a reductive subgroup has automorphic support. Clozel has conjectured a qualitative refinement of this result, which was first verified and quantified in the GLn case by Venkatesh. In this talk I will describe my thesis which extended this result to classical groups.

Deep Learning has always been divided into two phases: Training and Inference. The common practice for Deep Learning is training big networks on huge datasets. While very successful, such networks are only applicable to the type of data they were trained for and require huge amounts of annotated data, which in many cases are not available. In my thesis (guided by Prof. Irani), I invented ``Deep Internal Learning''. Instead of learning to generally solve a task for all inputs, we perform ``ad hoc'' learning for specific input. We train an image-specific network, we do it at test-time and on the test-input only, in an unsupervised manner (no label or ground-truth). In this regime, training is actually a part of the inference, no additional data or prior training is taking place. I will demonstrate how we applied this framework for various challenges: Super-Resolution, Segmentation, Dehazing, Transparency-Separation, Watermark removal. I will also show how this approach can be incorporated to Generative Adversarial Networks by training a GAN on a single image. If time permits I will also cover some partially related works.

Links to papers:
http://www.wisdom.weizmann.ac.il/~vision/zssr
http://www.wisdom.weizmann.ac.il/~vision/DoubleDIP
http://www.wisdom.weizmann.ac.il/~vision/ingan
http://www.wisdom.weizmann.ac.il/~vision/kernelgan
https://semantic-pyramid.github.io/
https://arxiv.org/abs/2006.11120
https://arxiv.org/abs/2103.15545

Near the ocean surface, mixing and turbulence modulate the transfer of heat, momentum, carbon and other properties, between the atmosphere and ocean interior. Accurate representation of these processes in General Circulation Models (GCMs) is crucial for simulating atmosphere-ocean interactions. However, all of these processes, generally known as boundary layer turbulence and submesoscale mixing, are on scales smaller than the grid used in GCMs, even at the highest possible resolution. Current submesoscale parameterizations represent the bulk of mixing developed across submesoscale fronts– the sharp interface between waters of different densities– but it has been shown to be too simplistic and unfitting in many circumstances. The presence of turbulence has been missing from these dynamics, and in this talk I will discuss the long-lasting problem of how to correctly include them.

Building toward a more complete understanding of these processes, a theoretical approach of perturbation analysis is used to include the effects of turbulence as a correction to classic frontogenesis (frontal sharpening) theory. This approach is next extended into a more realistic environment, using a suite of high resolution, turbulence resolving, numerical simulations. It is found that a variety of turbulent processes resulting from winds, waves, convection, and instabilities affect the formation of fronts. Furthermore, this analysis exposes severe limitations in existing techniques to predict potential vorticity dynamics in highly turbulent regimes. Lastly, I will discuss modifying the submesoscale parameterization in GCMs to represent the complex interactions with boundary layer turbulence.