Tim Marrinan, PhD

Data Scientist
Mathematics, Statistics, and Data Science Group
National Security Directorate
Pacific Northwest National Laboratory

I am an applied mathematician focused on bridging computational mathematics and rigorous machine learning. I am working to support the mental health of students in math and engineering, and advocating for students from historically excluded groups.

My Erdős number is 4

Education and Professional Experience

2021-2022: Postdoctoral Research Fellow (Prof. Fu’s research group, Oregon State University)
2019-2021: Postdoctoral Research Associate (COLORAMAP Group, Université de Mons )
2017-2019: Postdoctoral Research Associate (Signal & System Theory Group, Universität Paderborn)
2013-2017: PhD in Mathematics (Pattern Analysis Lab, Colorado State University)
2010-2013: MS in Mathematics (Pattern Analysis Lab, Colorado State University)
2004-2008: BA in Applied Mathematics and Geology (Whitman College)

Recent News:

[2023.10.20] Congratulations Prof. Shahana Ibrahim! Dr. Ibrahim recently accepted a tenure-track position with the joint appointment in the Departments of ECE and CS at the University of Central Florida as part of their new AI Initiative. You can find more details of her research on her website: (https://shahanaibrahimosu.github.io/)
[2023.08.22] New paper: Daniel Grey Wolnick, Dr. Shahana Ibrahim, Prof. Xiao Fu and I wrote a paper called, “Deep Learning from Noisy Labels via Robust Nonnegative Matrix Factorization-Based Design,” which was recently accepted to the 2023 CAMSAP conference. This is the first publication for Grey, which came from the research he did during his REU and Honor’s thesis with Prof. Fu. Congratulations, Grey! It is not common for an undergrad to get a first-author publication.
[2023.07.03] New paper: Maurice Kuschel, Dr. Tanuj Hasija, and I wrote a new paper titled, “Geodesic-based relaxation for deep canonical correlation analysis,” which has been accepted to the the IEEE Machine Learning for Signal Processing workshop (https://2023.ieeemlsp.org/) in Rome, Italy. I think this work is quite fun, and it will be presented in our special session on multiview representation learning. If you are also attending MLSP, swing by and say hi! The preprint and code will be forthcoming.
[2023.06.21] Upcoming workshop: Together with Dr. Tanuj Hasija, I will be hosting a special session on “Multiview representation learning for machine learning and data fusion,” at the IEEE Machine Learning for Signal Processing workshop (https://2023.ieeemlsp.org/) in Rome, Italy. There is a wonderful group of participants and we are looking forward to some lively discussions.
[2023.06.21] Congratulations to Prof. Xiao Fu! Prof. Fu was granted tenure and promoted to Associate Professor at Oregon State University following the end of the Spring Term 2023. It is a well-earned acknowledgement of the great work he has done in research, teaching, and mentoring since joining OSU.
[2023.06.21] Congratulations to Grey Wolnick! Grey worked with Prof. Fu, Shahana Ibrahim, and myself to develop an Honor’s Thesis on the topic of ‘Noisy Label Learning’. His thesis presentation was wonderful, and I am proud of the work he has done during his time with the group. Following his graduation
New paper: Dr. Hasija and I wrote a new paper titled, “A GLRT for estimating the number of correlated components in sample-poor mCCA,” which was recently accepted to the European Signal Processing Conference (EUSIPCO) in Serbia this year. If you are also attending, swing by and say hi! The preprint and code will be forthcoming.Grey will be joining a start-up in the bay area.
Top reviewer at AISTATS 2023: I am grateful to have been chosen by the Chairs of the AISTATS 2023 conference as one of the top reviewers this year, for the second year in a row. Top reviewers were selected based on the feedback received from the Area Chairs and comprise the top-10% of AISTATS reviewers. The list of top reviewers is available here: https://aistats.org/aistats2023/reviewers
[2022.12.05] I have accepted a new position! In early December I will join the Math, Stats, and Data Science Group of the National Security Directorate at the Pacific Northwest National Laboratory in Seattle, Washington. I am really excited to see what new problems they are working on.
Congratulations to Dr. Qi Lyu! Dr. Lyu recently defended his dissertation, titled, “Unsupervised Neural Representation Learning Through The Lens Of Nonlinear Mixture Identification” at Oregon State University under the supervision of Xiao Fu. In July he will be joining the Google research team in California. You can find details of his research at: https://web.engr.oregonstate.edu/~lyuqi/
EECS student success committee subgroup on mental health: I am excited to start working with students, faculty, and staff in the department of Electrical Engineering and Computer Science here at OSU to support the mental health of undergraduates. I hope to include some thoughts and experience on the topic on the teaching page in the coming months.
Congratulations to Prof. Xiao Fu! My current postdoctoral supervisor was recently award a prestigious NSF CAREER grant to fund his ongoing work in identifiable machine learning. Read the story here: https://engineering.oregonstate.edu/five-college-engineering-faculty-win-early-career-investigator-awards
[2022.05.20] Upcoming talk: I was kindly invited by Dr. Vikas Mishra to speak at the ‘International Control e-Seminar’ on May 20th at the Technische Universität Kaiserslautern. The talk will be available virtually. Details can be found here: https://www.mv.uni-kl.de/mec/events
Research experience for undergrads: Prof. Fu and I will be organizing a summer research experience for undergrads based on a couple of projects in machine learning and applied linear algebra. We are very excited to get started with our new cohort of students.
New paper: Dr. Hasija and I wrote a new paper titled, “A GLRT for estimating the number of correlated components in sample-poor mCCA,” which was recently accepted to the European Signal Processing Conference (EUSIPCO) in Serbia this year. If you are also attending, swing by and say hi! The preprint and code will be forthcoming.
Congratulations to Dr. Nicolas Nadisic! My good friend and colleague Nicolas has recently defended his dissertation, titled, “Sparsity and Nonnegativity in Least Squares Problems and Matrix Factorizations” at Université de Mons under the supervision of Nicolas Gillis. You can find his dissertation and wonderful code on his personal webpage: http://nicolasnadisic.xyz/
There have been lots of updates recently, so the above items are not necessarily in chronological order. Sorry, not sorry.
Top reviewer at AISTATS 2022: I am grateful to have been acknowledged by the Chairs of the AISTATS 2022 conference as one of the top reviewers this year. Top reviewers were selected based on the feedback received from the Area Chairs and comprise the top-10% of AISTATS reviewers. The list of top reviewers is available here: https://virtual.aistats.org/Conferences/2022/Reviewers
New research/travel grant: My collaborator Dr. Tanuj Hasija and I have received a research award from Paderborn University in honor of the University’s 50th anniversary. I will be visiting Dr. Hasija in May 2022 to work on our project, “Nonlinear disentanglement from multiple views: Identifiability and interpretability using deep learning-based solutions.”
I have changed positions! I am very excited to join the research group of Professor Xiao Fu in the Department of Electrical Engineering and Computer Science at Oregon State University. While in Corvallis, I plan to focus on identifiability conditions for machine learning algorithms, while continuing my ongoing research projects in nonnegative matrix factorization, multiset correlation analysis, and manifold optimization.
[2021.11.06] Python package released: Praneeth Balakrishna, Tanuj Hasija, and I released a Python package today for computing multiset correlation analysis on high dimensional data. This package translates and unifies some of our existing Matlab code, and is available on Github (https://github.com/praneeth-b/Correlation-Analysis-in-High-Dimensional-Data)
The published version of our paper, On a minimum enclosing ball of a collection of linear subspaces, is now available online via Linear Algebra and Its Applications.
SIAM Linear Algebra Conference: Next week (May 17th – 21st) is the SIAM Linear Algebra Conference, which was originally slated to be held in New Orleans. Unfortunately it will be organized virtually, but I will give a talk on Thursday about our work on practical verification of the Sufficiently Scattered Conditions for NMF identifiability.
Upcoming talks: I will be giving two talks in January. On January 6th I will speak about verifying sufficient conditions for minimum volume nonnegative matrix factorization at the Joint Math Meetings, and on January 21st I will speak about minimax nonnegative matrix factorization during EUSIPCO 2020.
Congratulations to Prof. Nicolas Gillis: My postdoctoral supervisor has just published his first book, Nonnegative Matrix Factorization. The book is available for purchase from SIAM, see the book page for details.
Our article, “On a minimum enclosing ball of a collection of linear subspaces” was accepted for publication in Linear Algebra and Its Applications: You can find a preprint PDF on the Publications page.
New article from Sarah Marrinan, MS in Marine Fisheries Review: My talented sister just published a retrospective on the Individual Fishing Quota Program in the quarterly journal of the National Oceanic and Atmospheric Administration (NOAA). You can find an open-access PDF on their site.
Congratulations to Dr. Andersen Man Shun Ang: My colleague Andersen recently passed his dissertation defense. He is the first PhD student of Prof. Nicolas Gillis to earn his doctorate. You can watch a video of his public defense or view his wonderful dissertation on his personal webpage: https://angms.science/
New code package available on the Research page: The MATLAB package minimax NMF code is now available on Research page. Given a set of matrices X_i (1≤i≤n) with the same number of rows, it computes W≥0 and H_i≥0 to minimize max1≤i≤n ||X_i – WH_i||_F^2 + λ logdet(WTW) where the second term is a volume regularizer. It outperforms existing methods in estimating endmembers (columns of W) that are only present in a small percentage of the samples.
New preprint available on the Publications page: “Hyperspectral unmixing with rare endmembers via minimax nonnegative matrix factorization,” with Nicolas Gillis was accepted to EUSIPCO. We will present the paper in Amsterdam this coming January.
COLORAMAP Day 2020: The yearly workshop of the COLORAMAP group was held on June 29th at UMons. We shared some lovely lectures with a variety of perspectives on nonnegative matrix factorization. Check out the updated group picture on Nicolas’ website
COLORAMAP Reading Group kickoff: In March, at the start of the coronavirus lockdown in Belgium, the group launched an online reading group to keep in contact with each other while working from home, and also to expand the breadth of our math reading. If anyone is interested in joining, please send me an email.

If you have any questions, please don’t hesitate to email me:
timothy.marrinan@pnnl.gov