Lizzie Buchanan
GPC Liaison Info Session
|
|
I will begin with a brief overview of the history of the GPC Liaison position, and describe what it is currently. Then we'll open it up for comments/questions about the past and present, and future of the position. The purpose of this info session is 55% for grad students interested in running for Liaison to learn more about it, and 45% for all grad students to talk about what they would like the Liaison position to be.
|
|
Jonathan Lindbloom
Hierarchical Bayesian Inverse Problems
|
|
This talk will be an introductory survey of theory and computational methods for Bayesian and hierarchical Bayesian inverse problems. Emphasizing visual illustrations, we will explore topics such as MAP estimation, hierarchical modelling, probabilistic sampling, and uncertainty quantification. We will conclude with a discussion of our recent work on solvers for a large class of problems.
|
|
Ryan Maguire
What You Don't Learn in Calculus: Pade Approximants and the Remez Exchange Algorithm
|
|
Given an analytic function, how might you approximate
it with a polynomial? The Taylor / Maclaurin series is
the safe bet, but are there better methods? What about
approximating with a rational function? In this talk
we'll discuss the Remez exchange algorithm for
computing the minimax polynomial, as well as Pade
approximants for the "best" rational approximation,
and how to compute it.
|
|
Steve Fan
Quadratic Reciprocity via Linear Algebra
|
|
Given two distinct odd primes $p$ and $q$, the law of quadratic reciprocity reveals a fascinating connection between the solvability of the quadratic equation $x^2=p$ in $\mathbb{F}_q$ and the solvability of $x^2=q$ in $\mathbb{F}_p$. This celebrated result, which was stated by Euler and Legendre and first proved by Gauss, now has over 240 published proofs and a variety of generalizations. In this talk, I will present an elegant proof of the law of quadratic reciprocity using linear algebra, essentially due to Schur.
|
|
Grant Molnar
Geometric Deep Learning
|
|
In the last two decades, deep learning has transformed from an academic curiosity into a paradigm-changing technology. Effective deep learning depends not only on a large corpus of training materials, but on choosing a network architecture appropriate to a given application. Although architecture choice has a massive impact on the effectiveness of deep learning, machine learning engineers often make this selection in ad hoc ways. Geometric deep learning provides a principled method for designing and implementing network architectures in a way that reflects the underlying data and the features we wish to extract from it. In this talk, we give a bare bones summary on what deep learning is and how it works, and then give an overview of geometric deep learning and how it can help us make more effective deep learning networks. This talk is inspired by "Geometric Deep Learning: Grids, Groups, Graphs, Geodesics, and Gauges" by Michael Bronstein, Joan Bruna, Taco Cohen, and Petar Velickovic.
|
|
Matthew Ellison
Teach Your Computer to Beat You at Games
|
|
How can we make a computer play a mean game of rock paper scissors, or chess, or texas hold'em? This will be an episodic GSS where we look at a game, discuss an algorithm to produce a good computer player, and then (prep permitting) face off against an implementation. I promise not to mention the prisoner's dilemma.
|
|
Ryan Maguire
Numerical Methods, Raytracing, and Newtonian Blackholes
|
|
Two things are true of classical mechanics:
it predicts the existence of black holes, and
there is no universal speed limit. In this talk
we'll neglect general relativity for the moment
and explore the world of "classical" black holes,
raytracing, and the numerical techniques used to
solve complex systems of motion.
|
|
David Freeman, Atticus McWhorter
Gold Q&A
|
|
On Friday February 24th, the paperwork and signed union cards were submitted to the National Labor Relations Board to request an election determining whether GOLD will be a federally recognized labor union or not. Though an official date has not been announced, the NLRB will soon be setting a date for the election, with the election likely taking place within the next month or so. In this GSS, we want to facilitate a discussion on the current state of the unionization process, what to expect before and during the election, and details on what exactly union recognition would mean and affect going forward. We may also discuss some possible ideas for how we can structure collective discussions internally within the math department.
|
|