Georgia Institute of Technology College of Sciences

Deep learning has achieved great success in recent years. One aspect overlooked by traditional deep-learning methods is uncertainty modeling, which can be very important in certain applications such as medical image classification and reinforcement learning. A standard way for uncertainty modeling is by adopting Bayesian inference. In this talk, I will share some of our recent work on scalable Bayesian inference by sampling, called optimal-transport sampling, motivated from the optimal-transport theory. Our framework formulates Bayesian sampling as optimizing a set of particles, overcoming some intrinsic issues of standard Bayesian sampling algorithms such as sampling efficiency and algorithm scalability. I will also describe how our sampling framework be applied to uncertainty and repulsive attention modeling in state-of-the-art natural-language-processing models.

https://bluejeans.com/197711728

I will compare and contrast a selection of popular equivalence relations on 4 manifolds, and explain some recent progress on classification results.

The speaker will hold online office hours from 3:00-4:00 pm for interested graduate students and postdocs.

In the last twenty or so years, a rich theory has emerged concerning combinatorial problems on infinite graphs endowed with extra structure, such as a topology or a measure. It turns out that there is a close relationship between this theory and distributed computing, i.e., the area of computer science concerned with problems that can be solved efficiently by a decentralized network of processors. In this talk I will outline this relationship and present a number of applications.
 

An infinite-type surface is a surface whose fundamental group is not finitely generated. These surfaces are “big,” having either infinite genus or infinitely many punctures. Recently, it was shown that mapping class groups of these infinite-type surfaces have a wealth of subgroups; for example, there are infinitely many surfaces whose mapping class group contains every countable group as a subgroup. By extending a theorem for finite-type surfaces to the infinite-type case — the Nielsen realization problem — we give topological obstructions to continuous embeddings of topological groups, with a few interesting examples.

 How can we recognize a map given certain combinatorial data? The Alexander method gives us the answer for self-homeomorphisms of finite-type surfaces. We can determine a homeomorphism of a surface (up to isotopy) based on how it acts on a finite number of curves. However, is there a way to apply this concept to recognize maps on other spaces? The answer is yes for a special class of maps, post-critically finite quadratic polynomials on the complex plane (Belk-Lanier-Margalit-Winarski). 

            In this talk, we will discuss Belk-Lanier-Margalit-Winarski’s methods, as well zome difficulties we face when trying to extend their methods to other settings.

Please join using the following link: MS Teams meeting

We shall discuss a recent paper of Wanless and Wood (arXiv:2008.00775), which proves a Lovász Local Lemma type result using inductive counting arguments.
For example, in the context of hypergraph colorings, under LLL-type assumptions their result typically gives exponentially many colorings (usually more than the textbook proof of LLL would give).
We will present a probabilistic proof of the Wanless-Wood result, and discuss some applications to k-SAT, Ramsey number lower bounds, and traversals, as time permits.