Past events

Data science seminars hosted by the The Institute for Mathematics and Its Applications (IMA) take place on Tuesdays from 1:25 p.m. - 2:25 p.m.

This week, Elizaveta Rebrova (University of California, Los Angeles), will be giving a lecture titled "Quantile-based Iterative Methods for Corrupted Systems of Linear Equations".

Registration is required to access the Zoom webinar.

Abstract

One of the most ubiquitous problems arising across the sciences is that of solving large-scale systems of linear equations Ax = b. When it is infeasible to solve the system directly by inversion, light and scalable iterative methods can be used instead, such as, Randomized Kaczmarz (RK) algorithm, or Stochastic Gradient Descent (SGD). The classical extensions of RK/SGD to noisy (inconsistent) systems work by showing that the iterations of the method still approach the least squares solution of the system until a certain convergence horizon (that depends on the noise size). However, in order to handle large, sparse, potentially adversarial corruptions, one needs to modify the algorithms to avoid corruptions rather than try to tolerate them -- and quantiles of the residual provide a natural way to do so. In this talk, I present QuantileRK and QuantileSGD, the versions of two classical iterative algorithms aimed at linear systems with adversarially corrupted vector b. Our methods work on up to 50% of incoherent corruptions, and up to 20% of adversarial corruptions (that consistently create an "alternative" solution of the system). Our theoretical analysis shows that under some standard assumptions on the measurement model, despite corruptions of any size, both methods converge to the true solution with exactly the same rate as RK on an uncorrupted system up to an absolute constant. Based on the joint work with Jamie Haddock, Deanna Needell, and Will Swartworth.

Biography

Liza Rebrova is currently a postdoctoral scholar at the Computational Research Division of the Lawrence Berkeley National Lab. From 2018 to the end of 2020, she worked as an Assistant Adjunct Professor at the UCLA Department of Mathematics (Computational and Applied Math Group, mentored by Professors Deanna Needell and Terence Tao). She received a Ph.D. in Mathematics from the University of Michigan in 2018 (advised by Prof. Roman Vershynin) and a Specialist degree from Moscow State University in 2012. Her research involves interactions with high-dimensional probability, random matrix theory, mathematical data science, and numerical linear algebra, with the main goal to study large high-dimensional data objects in the presence of randomness and to develop randomized algorithms that efficiently process complex data. She is a recipient of the Allen Shields Memorial Fellowship (UofMichigan, 2018) and postdoctoral sponsorship by Capital Fund Management (UCLA, 2018-2020). 

View the full list of IMA data science seminars.

Data science seminars hosted by the The Institute for Mathematics and Its Applications (IMA) take place on Tuesdays from 1:25 p.m. - 2:25 p.m.

This week, Ido Kaminer (Technion-Israel Institute of Technology), will be giving a lecture titled "The Ramanujan Machine: Using Algorithms for the Discovery of Conjectures on Mathematical Constants".

Registration is required to access the Zoom webinar.

Abstract

In the past, new conjectures about fundamental constants were discovered sporadically by famous mathematicians such as Newton, Euler, Gauss, and Ramanujan. The talk will present a different approach – a systematic algorithmic approach that discovers new mathematical conjectures on fundamental constants. We call this approach “the Ramanujan Machine”. The algorithms found dozens of well-known formulas as well as previously unknown ones, such as continued fraction representations of π, e, Catalan’s constant, and values of the Riemann zeta function. Part of the conjectures were in retrospect simple to prove, whereas others remained so far unproved. We will discuss these puzzles and wider open questions that arose from this algorithmic investigation – specifically, a newly-discovered algebraic structure that seems to generalize all the known formulas and connect between fundamental constants. We will also discuss two algorithms that proved useful in finding conjectures: a variant of the meet-in-the-middle algorithm and a gradient descent algorithm tailored to the recurrent structure of continued fractions. Both algorithms are based on matching numerical values; consequently, they conjecture formulas without providing proofs or requiring prior knowledge of the underlying mathematical structure. This way, our approach reverses the conventional usage of sequential logic in formal proofs; instead, using numerical data to unveil mathematical structures and provide leads to further mathematical research.

Biography

Ido Kaminer joined the Technion as an assistant professor and an Azrieli Faculty Fellow in 2018, after a postdoc at MIT as a Rothschild Fellow, MIT-Technion Fellow, and a Marie Curie Fellow. In his PhD, Ido discovered new classes of accelerating beams in nonlinear optics and electromagnetism, for which he received the 2012 Israel Physical Society Prize, and the 2014 APS (American Physical Society) Award for Outstanding Doctoral Dissertation in Laser Science. Ido was the first Israeli to ever win an APS award for his PhD thesis. He was chosen to the 2020 list of 40 promising leaders under 40 by TheMarker and won multiple awards and grants recently including the ERC Starting Grant, and the 2021 Blavatnik Award for Young Scientists in Israel.

View the full list of IMA data science seminars.

Data science seminars hosted by the The Institute for Mathematics and Its Applications (IMA) take place on Tuesdays from 1:25 p.m. - 2:25 p.m.

This week, John Wright (Columbia University), will be giving a lecture titled "Deep Networks and the Multiple Manifold Problem".

Registration is required to access the Zoom webinar.

Abstract

Data with low-dimensional nonlinear structure are ubiquitous in engineering and scientific problems. We study a model problem with such structure—a binary classification task that uses a deep fully-connected neural network to classify data drawn from two disjoint smooth curves on the unit sphere. Aside from mild regularity conditions, we place no restrictions on the configuration of the curves. We prove that when (i) the network depth is large relative to certain geometric properties that set the difficulty of the problem and (ii) the network width and number of samples is polynomial in the depth, randomly-initialized gradient descent quickly learns to correctly classify all points on the two curves with high probability. To our knowledge, this is the first generalization guarantee for deep networks with nonlinear data that depends only on intrinsic data properties. Our analysis draws on ideas from harmonic analysis and martingale concentration for handling statistical dependencies in the initial (random) network. We sketch applications to invariant vision, and to gravitational wave astronomy, where leveraging low-dimensional structure leads to statistically optimal tests for identifying signals in noise.

Biography

John Wright is an associate professor in Electrical Engineering at Columbia University. He is also affiliated with the Department of Applied Physics and Applied Mathematics and Columbia’s Data Science Institute. He received his PhD in Electrical Engineering from the University of Illinois at Urbana Champaign in 2009. Before joining Columbia he was with Microsoft Research Asia from 2009-2011. His research interests include sparse and low-dimensional models for high-dimensional data, optimization (convex and otherwise), and applications in imaging and vision. His work has received a number of awards and honors, including the 2012 COLT Best Paper Award and the 2015 PAMI TC Young Researcher Award. 

View the full list of IMA data science seminars.

Data science seminars hosted by the The Institute for Mathematics and Its Applications (IMA) take place on Tuesdays from 1:25 p.m. - 2:25 p.m.

This week, Jeffrey Rosenthal (University of Toronto), will be giving a lecture titled "Adapting the Metropolis Algorithm".

Registration is required to access the Zoom webinar.

Abstract

The Metropolis Algorithm is an extremely useful and popular method of approximately sampling from complicated probability distributions. "Adaptive" versions automatically modify the algorithm while it runs, to improve its performance on the fly, but at the risk of destroying the Markov chain properties necessary for the algorithm to be valid.  In this talk, we will illustrate the Metropolis algorithm using a very simple JavaScript example (http://probability.ca/jeff/js/metropolis.html).  We will then discuss adaptive MCMC, and present examples and theorems concerning its ergodicity and efficiency.

Biography

Jeffrey S. Rosenthal is a professor of Statistics at the University of Toronto, specializing in Markov chain Monte Carlo (MCMC) algorithms. He received his BSc from the University of Toronto at age 20, and his PhD in Mathematics from Harvard University at age 24. He was awarded the 2006 CRM-SSC Prize, the 2007 COPSS Presidents' Award, the 2013 SSC Gold Medal, and fellowship of the Institute of Mathematical Statistics and of the Royal Society of Canada. He has published well over one hundred research papers, and five books (including the Canadian bestseller Struck by Lightning: The Curious World of Probabilities). His web site is www.probability.ca, and on Twitter he is @ProbabilityProf.

View the full list of IMA data science seminars.

Data science seminars hosted by the The Institute for Mathematics and Its Applications (IMA) take place on Tuesdays from 1:25 p.m. - 2:25 p.m.

This week, Faming Liang (Purdue University), will be giving a lecture titled "Consistent Sparse Deep Learning: Theory and Computation".

Registration is required to access the Zoom webinar.

Abstract

Deep learning has been the engine powering many successes of data science. However, the deep neural network (DNN), as the basic model of deep learning, is often excessively over-parameterized, causing many difficulties in training, prediction and interpretation. We propose a frequentist-like method for learning sparse DNNs and justify its consistency under the Bayesian framework: the proposed method could learn a sparse DNN with at most O(n/log(n))O(n/log⁡(n)) connections and nice theoretical guarantees such as posterior consistency, variable selection consistency and asymptotically optimal generalization bounds. In particular, we establish posterior consistency for the sparse DNN with a mixture Gaussian prior, show that the structure of the sparse DNN can be consistently determined using a Laplace approximation-based marginal posterior inclusion probability approach, and use Bayesian evidence to elicit sparse DNNs learned by an optimization method such as stochastic gradient descent in multiple runs with different initializations. The proposed method  is computationally more efficient than standard Bayesian methods for large-scale sparse DNNs.  The numerical results indicate that the proposed method can perform very well for large-scale network compression and high-dimensional nonlinear variable selection, both advancing interpretable machine learning.  The talk is based on a joint work with Yan Sun and Qifan Song.

Biography

Faming Liang is Professor of Statistics at Purdue University. Before joining Purdue, he held a faculty position at University of Florida and Texas A&M University. Faming has wide research interests, including machine learning, Monte Carlo methods, bioinformatics, high-dimensional statistics, and big data. He is ASA fellow and IMS fellow, and has published over 120 journal papers.

View the full list of IMA data science seminars.

Data science seminars hosted by the The Institute for Mathematics and Its Applications (IMA) take place on Tuesdays from 1:25 p.m. - 2:25 p.m.

This week, Jianfeng Lu (Duke University), will be giving the lecture

View the full list of IMA data science seminars.

Data science seminars hosted by the The Institute for Mathematics and Its Applications (IMA) take place on Tuesdays from 1:25 p.m. - 2:25 p.m.

This week, Facundo Mémoli (The Ohio State University), will be giving the lecture

View the full list of IMA data science seminars.

Data science seminars hosted by the The Institute for Mathematics and Its Applications (IMA) take place on Tuesdays from 1:25 p.m. - 2:25 p.m.

This week, Kenth Monsen (Telenor Research), will be giving a lecture titled "Using Telco Data to Fight Epidemics".

Abstract

In this talk we will discuss telecom data to estimate human mobility at country-wide scales, and the utilization of such data to better understand the spread of infectious diseases like dengue, malaria and covid-19. This will further be complemented with insights and experiences gathered on privacy, data security, and the need to align solutions with national public health initiatives.

Biography

Kenth Engø-Monsen, PhD, is a senior research scientist and data scientist in Telenor Research. He is currently leading Telenor Group’s initiative on big data for social good. With more than 15 years of experience in telecom, Dr. Engø-Monsen has extensive knowledge in the field of telecom data, social network analysis, and applied research using mobile data. He is the co-inventor on numerous patents, has published numerous academic papers in mathematics, computer science, data science, and social science. He received his Master’s in 1995 in Industrial Mathematics from NTNU, Trondheim, Norway, and PhD in 2000 in Computer Science from University of Berge, Norway. 

View the full list of IMA data science seminars.

Save the date for the Spring 2021 CSE Virtual Career Fair. Mark your calendar for Tuesday, February 9, 2021 and start preparing over winter break.

Spring 2021 CSE Virtual Career Fair
Tuesday, February 9, 2021
11 a.m.-6 p.m. Central Time
Virtual fair platform: Career Fair Plus

Similar to the Fall 2020 fair, the Spring fair will be held via the virtual platform, Career Fair Plus. The platform provides features that will allow you to connect personally with employers.

Virtual Career Fair features

• No waiting in line! Students can pre-register for a time to speak with employers to video chat.
• Speak with employers from the comfort and safety of your home.
• Research employers prior to the fair, as well as during the fair by entering the employer's group chat and info session room.

Download the Career Fair Plus app

Download the Career Fair Plus App in the App Store for Apple devices or the Google Play Store for Android devices. Once downloaded, search for the University of Minnesota, then College of Science and Engineering Career Fair. Alternatively, you can view the event online.

You can use the app now to begin researching employers. Winter break is a great time to begin researching employers, and new employers are being added each day! Additional instructions on how to sign-up to speak with employers will be provided soon after the beginning of spring semester.

Get your resume ready and prepare for interviews

Visit the CSE Career Fair website for resume and cover letter writing guides, interviewing tips, advice from employers, and more.

Career counselors are also available over winter break for individual appointments. Please reach out by scheduling a virtual appointment or by sending an email to csecareer@umn.edu.

Prospective students can RSVP for an information session to learn about the following graduate programs:

• Computer Science M.S.
• Computer Science MCS
• Computer Science Ph.D.
• Data Science M.S.
• Data Science Post-Baccalaureate Certificate

During the information session, we will go over the following:

• Requirements (general)
• Applying
• Prerequisite requirements
• What makes a strong applicant
• Funding
• Resources
• Common questions
• Questions from attendees