Past events

The UMN Machine Learning Seminar Series brings together faculty, students, and local industrial partners who are interested in the theoretical, computational, and applied aspects of machine learning, to pose problems, exchange ideas, and foster collaborations. The talks are every Thursday from 12 p.m. - 1 p.m. during the Fall 2021 semester.

This week's speaker, Mikhail Belkin (University of California San Diego), will be giving a talk titled "The Polyak-Lojasiewicz condition as a framework for over-parameterized optimization and its application to deep learning."

Abstract

The success of deep learning is due, to a large extent, to the remarkable effectiveness of gradient-based optimization methods applied to large neural networks. In this talk I will discuss some general mathematical principles allowing for efficient optimization in over-parameterized non-linear systems, a setting that includes deep neural networks. I will discuss that optimization problems corresponding to these systems are not convex, even locally, but instead satisfy the Polyak-Lojasiewicz (PL) condition on most of the parameter space, allowing for efficient optimization by gradient descent or SGD. I will connect the PL condition of these systems to the condition number associated with the tangent kernel and show how a non-linear theory for those systems parallels classical analyses of over-parameterized linear equations. As a separate related development, I will discuss a perspective on the remarkable recently discovered phenomenon of transition to linearity (constancy of NTK) in certain classes of large neural networks. I will show how this transition to linearity results from the scaling of the Hessian with the size of the network controlled by certain functional norms. Combining these ideas, I will show how the transition to linearity can be used to demonstrate the PL condition and convergence for a general class of wide neural networks. Finally I will comment on systems which are ''almost'' over-parameterized, which appears to be common in practice.

Biography

Mikhail Belkin received his Ph.D. in 2003 from the Department of Mathematics at the University of Chicago. His research interests are in theory and applications of machine learning and data analysis. Some of his well-known work includes widely used Laplacian Eigenmaps, Graph Regularization and Manifold Regularization algorithms, which brought ideas from classical differential geometry and spectral analysis to data science. His recent work has been concerned with understanding remarkable mathematical and statistical phenomena observed in deep learning. This empirical evidence necessitated revisiting some of the basic concepts in statistics and optimization. One of his key recent findings is the "double descent" risk curve that extends the textbook U-shaped bias-variance trade-off curve beyond the point of interpolation. Mikhail Belkin is a recipient of a NSF Career Award and a number of best paper and other awards. He has served on the editorial boards of the Journal of Machine Learning Research, IEEE Pattern Analysis and Machine Intelligence and SIAM Journal on Mathematics of Data Science.

Tuesday, September 21 and Wednesday, September 22, 2021
Noon - 6 p.m. each day
The fair will be held via Handshake

View the day one list of companies recruiting and the day two list of companies recruiting now and begin signing up for time slots to speak individually with companies beginning September 14, 2021 at 8:00 a.m.

Visit the Career Information for Students webpage for more information!

For questions, contact the CSE Career Center at csecareer@umn.edu or by calling 612-624-4090.
 

The last day to apply for fall undergraduate graduation is Tuesday, September 21.

View the full academic schedule on One Stop.
 

The Institute for Mathematics and Its Applications (IMA) Data Science Seminars are a forum for data scientists of IMA academic and industrial partners to discuss and learn about recent developments in the broad area of data science. The seminars take place on Tuesdays from 1:25 p.m. - 2:25 p.m.

This week's speaker, Sara Algeri (University of Minnesota), will be giving a talk titled "Handling model uncertainties via informative Goodness-of-Fit."

Abstract

When searching for signals of new astrophysical phenomena, astrophysicists have to account for several sources of non-random uncertainties which can dramatically compromise the sensitivity of the experiment under study. Among these, model uncertainty arising from background mismodeling is particularly dangerous and can easily lead to highly misleading results. Specifically, overestimating the background distribution in the signal region increases the chances of falsely rejecting the hypothesis that the new source is present. Conversely, underestimating the background outside the signal region leads to an artificially enhanced sensitivity and a higher likelihood of claiming a false discovery. The aim of this work is to provide a self-contained framework to perform modeling, estimation, and inference under background mismodeling. The method proposed allows incorporating the (partial) scientific knowledge available on the background distribution, and provides a data-updated version of it in a purely nonparametric fashion, and thus, without requiring the specification of prior distributions. If a calibration (or control regions) is available, the solution discussed does not require the specification of a model for the signal, however when available, it allows to further improve the accuracy of the analysis and to detect additional and unexpected signal sources.

Biography

Sara Algeri has been an Assistant Professor in the School of Statistics at the University of Minnesota since August 2018. Her appointment at UMN started soon after completing her doctoral studies in statistics at Imperial College London (UK). Her research interests mainly lie in astrostatistics, computational statistics, and statistical inference. The main purpose of her work is to provide generalizable statistical solutions which directly address fundamental scientific questions, and can at the same time be easily applied to any other scientific problem following a similar statistical paradigm. In line with this, motivated by the problem of the detection of particle dark matter, her current research focuses on statistical inference for signal detection under lack of regularity. She is also interested in uncertainty quantification in the context of astrophysical discoveries.

Tuesday, September 21 and Wednesday, September 22, 2021
Noon - 6 p.m. each day
The fair will be held via Handshake

View the day one list of companies recruiting and the day two list of companies recruiting now and begin signing up for time slots to speak individually with companies beginning September 14, 2021 at 8:00 a.m.

Visit the Career Information for Students webpage for more information!

For questions, contact the CSE Career Center at csecareer@umn.edu or by calling 612-624-4090.
 

The last day to cancel full semester classes and not receive a "W" is Monday, September 20. This is also the last day to receive a 75% tuition refund for canceling full semester classes.

In addition, this is the last day to add classes without college approval and to change grade basis (A-F or S/N) for full semester classes.

View the full academic schedule on One Stop.
 

Are you a current CSE undergraduate student interested in doing high school outreach to local high school 11th/12th graders? Please consider the brand new Rooted in STEM program!

Information Session for Potential Mentors
Friday, September 17
4 - 5 p.m.
Bruininks Hall 420B

About the Program
This program will focus on historically excluded and underserved students from Twin Cities high schools who are interested in STEM and will be capped at 25 participants. The ratio for mentors/mentees will likely be one undergraduate mentor to every three high school mentees.

Undergraduates with BIPOC, First Generation, and LGBTQIA+ identities are especially encouraged to apply!

Requirements for Mentors
Mentors will be asked to mentor one Saturday per month (October-April) for 4 hours each in addition to several pre-program trainings.

To support 25 high school participants, Rooted in STEM is now recruiting 8-9 CSE undergraduate students with a 2.5+ cumulative GPA to serve as mentors. Highly dedicated mentors will be essential program staff and must be able to fulfil to the following:

Minimum Commitment

• Pass a background check required by the University’s Safety of Minors policy (For more information, please visit the policy website: policy.umn.edu/operations/minorsafety)
• Attend required mentor training on EITHER Friday, October 8 from 8:00-10:00am OR Wednesday, October 13 from 11:30am-1:00pm.
• Attend all program sessions from 9:30am-1:30pm on the following Saturdays: October 16, November 13, December 11, January 22, February 19, March 19, and April 30
• Attend periodic mentor meetings

Perks

• Mentors who fulfill their commitment will be eligible for a $500 scholarship at the end of the program
• Catered lunch provided during Saturday program sessions

Questions? Email Dan Garrison, Assistant Director for Diversity and Inclusion for CSE Collegiate Life.

In collaboration with the Minnesota Center for Industrial Mathematics, the Industrial Problems Seminars are a forum for industrial researchers to offer a first-hand glimpse into industrial research. The seminars take place Fridays from 1:25 p.m. - 2:25 p.m.

This week will be a SIAM Internship Panel.

Come learn about the process of finding, interviewing, and getting jobs in industry! Panelists Brendan Cook, Jacob Hegna, Drisana Mosaphir, Cole Wyeth, and Amber Yuan will be here to answer all your questions about finding and participating in internships both before and during the pandemic.

The UMN Machine Learning Seminar Series brings together faculty, students, and local industrial partners who are interested in the theoretical, computational, and applied aspects of machine learning, to pose problems, exchange ideas, and foster collaborations. The talks are every Thursday from 12 p.m. - 1 p.m. during the Fall 2021 semester.

This week's speaker, Holger Roth (NVIDIA), will be giving a talk titled "Tackling the Challenges of Next-generation Healthcare: NVIDIA’s Applied Research in Medical Imaging."

Abstract

Recent advances in computer vision and artificial intelligence caused a paradigm shift in medical image computing and radiological image analysis. Deep learning has been widely applied to many radiological applications, replacing, or working together with conventional methods. The advantage of being able to learn from that data directly is promising for many imaging tasks. Some key factors and current challenges preventing the widespread adaption of machine learning techniques in the clinic are algorithmic considerations, computational power, and, most critically, high-quality data for training.

NVIDIA wants to provide solutions to make the widespread adoption of deep learning and artificial intelligence easier in the real world. This talk will highlight NVIDIA’s efforts in the healthcare sector and medical imaging research, for example, around federated learning and COVID-19 image analysis, and introduce platforms & hardware considerations for modern machine learning at scale.

Biography

Holger Roth is a Sr. Applied Research Scientist at NVIDIA focusing on deep learning for medical imaging. He has been working closely with clinicians and academics over the past several years to develop deep learning based medical image computing and computer-aided detection models for radiological applications. He is an Associate Editor for IEEE Transactions of Medical Imaging and holds a Ph.D. from University College London, UK. In 2018, he was awarded the MICCAI Young Scientist Publication Impact Award.

The last day to receive a 100% tuition refund for canceling full semester classes is Tuesday, September 14.

View the full academic schedule on One Stop.