Seminars and Colloquia by Series

Convolutional Neural Network with Structured Filters

Series
Applied and Computational Mathematics Seminar
Time
Monday, April 16, 2018 - 13:55 for 1 hour (actually 50 minutes)
Location
Skiles 005
Speaker
Xiuyuan ChengDuke University
Filters in a Convolutional Neural Network (CNN) contain model parameters learned from enormous amounts of data. The properties of convolutional filters in a trained network directly affect the quality of the data representation being produced. In this talk, we introduce a framework for decomposing convolutional filters over a truncated expansion under pre-fixed bases, where the expansion coefficients are learned from data. Such a structure not only reduces the number of trainable parameters and computation load but also explicitly imposes filter regularity by bases truncation. Apart from maintaining prediction accuracy across image classification datasets, the decomposed-filter CNN also produces a stable representation with respect to input variations, which is proved under generic assumptions on the basis expansion. Joint work with Qiang Qiu, Robert Calderbank, and Guillermo Sapiro.

Chaotic Transition States on the Monkey Saddle

Series
CDSNS Colloquium
Time
Monday, April 16, 2018 - 11:15 for 1 hour (actually 50 minutes)
Location
skiles 005
Speaker
Thomas BartschLoughborough University

Please Note: Transition State Theory describes how a reactive system crosses an energy barrier that is marked by a saddle point of the potential energy. The transition from the reactant to the product side of the barrier is regulated by a system of invariant manifolds that separate trajectories with qualitatively different behaviour. The situation becomes more complex if there are more than two reaction channels, or possible outcomes of the reaction. Indeed, the monkey saddle potential, with three channels, is known to exhibit chaotic dynamics at any energy. We investigate the boundaries between initial conditions with different outcomes in an attempt to obtain a qualitative and quantitative description of the relevant invariant structures.

TBA

Trace Test

Series
Student Algebraic Geometry Seminar
Time
Friday, April 13, 2018 - 10:00 for 1 hour (actually 50 minutes)
Location
Skiles 006
Speaker
Tim DuffGeorgia Tech
The fundamental data structures for numerical methods in algebraic geometry are called "witness sets." The term "trace test" refers to certain numerical methods which verify the completeness of such witness sets. It is natural to ask questions about the complexity of such a test and in what sense its output may be regarded as "proof." I will give a basic exposition of the trace test(s) with a view towards these questions

Quenched survival of Bernoulli percolation on Galton-Watson trees

Series
Stochastics Seminar
Time
Thursday, April 12, 2018 - 15:05 for 1 hour (actually 50 minutes)
Location
Skiles 006
Speaker
Joshua RosenbergUniversity of Pennsylvania
In this talk I will explore the subject of Bernoulli percolation on Galton-Watson trees. Letting $g(T,p)$ represent the probability a tree $T$ survives Bernoulli percolation with parameter $p$, we establish several results relating to the behavior of $g$ in the supercritical region. These include an expression for the right derivative of $g$ at criticality in terms of the martingale limit of $T$, a proof that $g$ is infinitely continuously differentiable in the supercritical region, and a proof that $g'$ extends continuously to the boundary of the supercritical region. Allowing for some mild moment constraints on the offspring distribution, each of these results is shown to hold for almost surely every Galton-Watson tree. This is based on joint work with Marcus Michelen and Robin Pemantle.

The extremal functions for triangle-free graphs with excluded minors

Series
Graph Theory Seminar
Time
Thursday, April 12, 2018 - 13:30 for 1 hour (actually 50 minutes)
Location
Skiles 005
Speaker
Youngho YooMath, GT
A classic theorem of Mader gives the extremal functions for graphs that do not contain the complete graph on p vertices as a minor for p up to 7. Motivated by the study of linklessly embeddable graphs, we present some results on the extremal functions of apex graphs with respect to the number of triangles, and on triangle-free graphs with excluded minors. Joint work with Robin Thomas.

An upper bound on the smallest singular value of a square random matrix

Series
Analysis Seminar
Time
Wednesday, April 11, 2018 - 13:55 for 1 hour (actually 50 minutes)
Location
Skiles 005
Speaker
Kateryna TatarkoUniversity of Alberta
Consider an n by n square matrix with i.i.d. zero mean unit variance entries. Rudelson and Vershynin showed that its smallest singular value is bounded from above by 1/sqrt{n} with high probability, under the assumption of the bounded fourth moment of the entries. We remove the assumption of the bounded fourth moment, thereby extending the result of Rudelson and Vershynin to a wide range of distributions.

What is Weak KAM Theory?

Series
Research Horizons Seminar
Time
Wednesday, April 11, 2018 - 12:10 for 1 hour (actually 50 minutes)
Location
Skiles 006
Speaker
Albert FathiGeorgia Tech
The goal of this lecture is to explain and motivate the connection between Aubry-Mather theory (Dynamical Systems), and viscosity solutions of the Hamilton-Jacobi equation (PDE).This connection is the content of weak KAM Theory.The talk should be accessible to the “generic” mathematician. No a priori knowledge of any of the two subjects is assumed.The set-up of this theory is classical mechanical systems, in its Lagrangian formulation to take advantage of the action principle. This is the natural setting for Celestial Mechanics. Today it is also the setting for motions of satellites in the solar system.Hamilton found a reformulation of Lagrangian mechanics in terms of position and momentum instead of position and speed. In this formulation appears the Hamilton-Jacobi equation. Although this is a partial differential equation, its solutions allow to find solutions of the Hamiltonian (or Lagrangian) systems which are, in fact, governed by an ordinary differential equation.KAM (Kolmogorov-Arnold-Moser) theorem addressed at its beginning (Kolomogorov) the problem of stability of the solar system. It came as a surprise, since Poincare ́’s earlier work pointed to instability. In fact, some initial conditions lead to instability (Poincare ́) and some others lead to stability(Kolomogorov).Aubry-Mather theory finds some more substantial stable motion that survives outside the region where KAM theorem applies.The KAM theorem also provides global differentiable solutions to the Hamilton-Jacobi equation.It is known that the Hamilton-Jacobi equation usually does not have smooth global solutions. Lions & Crandall developed a theory of weak solutions of the Hamilton-Jacobi equation.Weak KAM theory explains how the Aubry-Mather sets can be obtained from the points where weak solutions of the Hamilton-Jacobi equation are differentiable.

L-infinity instability of Prandtl's layers

Series
PDE Seminar
Time
Tuesday, April 10, 2018 - 15:00 for 1 hour (actually 50 minutes)
Location
Skiles 006
Speaker
Toan NguyenPenn State University
In 1904, Prandtl introduced his famous boundary layer theory to describe the behavior of solutions of incompressible Navier Stokes equations near a boundary in the inviscid limit. His Ansatz was that the solution of Navier Stokes can be described as a solution of Euler, plus a boundary layer corrector, plus a vanishing error term in $L^\infty$. In this talk, I will present a recent joint work with E. Grenier (ENS Lyon), proving that, for a class of regular solutions of Navier Stokes equations, namely for shear profiles that are unstable to Rayleigh equations, this Prandtl's Ansatz is false. In addition, for shear profiles that are monotone and stable to Rayleigh equations, the Prandtl's asymptotic expansions are invalid.

IBM PonderThis monthly challenge

Series
Other Talks
Time
Tuesday, April 10, 2018 - 11:00 for 1 hour (actually 50 minutes)
Location
Skiles 006
Speaker
Prof. Oded MargalitCTO, IBM Cyber security center of excellence at Ben Gurion, University of the Negev

Please Note: [CV: Prof. Oded Margalit, PhD in Computer Science from Tel-Aviv University under the supervision of Prof. Zvi Galil has worked at IBM's Haifa research lab on machine learning, constraint satisfaction, verification and more. Currently he is the CTO of the IBM Cyber security center of excellence at Ben Gurion University of the Negev. Oded participates in organising several computer science competitions (like the international IEEEXtreme and the national CodeGuru). He loves riddles and authors the monthly challenge corner of IBM research: "Ponder-This".]

IBM research runs a mathematical challenge site. Every month a new challenge is posted; as well as a solution for the previous month's riddle. Prof. Oded Margalit is the puzzlemaster, for the last decade. In the talk, he will survey some of the riddles over the years, and tell some anecdotes about the challenges and the solvers. For example: A PRL paper born from a riddle on random walks; ITA-2014 paper on water hose model (using quantum entanglement to break location based encryption); Games: 2048, Kakuro, Infinite chess game, the probability of a backgammon to end with a double, Fisher Foul Chess and more. Minimal hash function, Combinatorial Test Design; A solver from Intensive Care Unit and other stories; Finding a natural number n such that round ((1+2 cos(20))^n) is divisible by 10^9; We'll leave you with a still open question about Permutation-firing cannon... Don't worry - no high math knowledge is assumed.

Pages