Wenjing Wang joined our team as a postdoc. She just graduated from Virginia Tech and will be working on the development of algorithms for large-scale computationally expensive black-box optimization problems and their application to high energy physics problems.
No OWG meeting this month (holiday)! Next one will be June 24.
Welcome Renaud and Jangho!
Renaud Saltet joined us as a summer intern on May 13. He will be supporting the Co-Optima project.
Jangho Park joined us as a postdoc on May 20. He will be working on our groundwater management project, focusing on optimization and uncertainty quantification tasks.
Two talks at SIAM CSE
Timur Takhtaganov presented research work on "Bayesian Inference Using Adaptive Gaussian Process Models with Application to Cosmology".
Juliane Mueller organized together with Matt Menickelly (ANL) a minisymposium on Derivative-free and Global Optimization. Juliane presented her recent work on Derivative-free Optimization of Computationally Expensive Functions with Noisy Data.
November 26, 11am, 59-4022: OWG meeting on Machine Learning
Reetik is joining us as a postdoc to work on a project related to groundwater management. He will develop a data-informed decision support tool to enable the sustainable management of California groundwater. Reetik received his Ph.D. from MIT's Department of Civil and Environmental Engineering.
Mueller invited to present her work on large-scale optimization at INFORMS Annual Meeting in Phoenix, AZ
Mueller was invited to discuss her recent work on large-scale black-box expensive optimization at the INFORMS Annual Meeting. She discussed the difficulties and opportunities of exploiting data analysis methods to solve large-dimensional optimization problems.
Yu-Hang Tang present in the OWG
A Practical Approach for Solving Positive Semidefinite Quadratic Programming Problems with Linear Equality Constraints
This is motivated by his work to improve molecular dynamics simulations.
Mike Minion presented his recent work in the Optimization Working Group
Mike Minion presented his work on parallel-in-time PDE-constrained optimization using PFASST - a collaborative work with Sebastian Gotschel at ZIB. More information on PFASST and Mike's research can be found here.
Two openings for postdoctoral researchers
The Center for Computational Sciences and Engineering has an immediate opening for two postdoctoral researchers to work on the development of computational optimization algorithms and decision support systems. The focus of the first position is the development of methods for large-scale computationally expensive optimization problems. The second position focuses on the development of data-informed surrogate models for decision support systems in the water-energy nexus.
We are looking for candidates with experience in nonlinear computational optimization including mixed-integer, derivative-free, or global optimization. Experience with surrogate models, data analysis, machine learning, and uncertainty quantification is a plus.
The candidates will work independently as well as collaboratively in a multidisciplinary team environment that includes mathematicians, computer/computational scientists, and domain scientists. The candidates will have the opportunity to apply their developments to exciting application areas.
For more information about these particular openings, please contact Juli Mueller at JulianeMueller (AT) lbl.gov
To apply, please go to the postings at
large scale optimization:
The posting shall remain open until the position is filled, however, for full consideration, please apply by August, 15, 2018.
Juli Mueller presents her recent research on sensitivity analysis informed optimization at the SIAM Annual Meeting in Portland
Juli organized a mini-symposium at the SIAM Annual Meeting whose topic was "Math Tools for Optimization, Uncertainty Quantification, and Sensitivity Analysis in Numerical Simulations". The presentations -- all by researchers from the DOE National Laboratories -- focused on PDE constrained optimization, the UQ Toolkit, managing ensembles of simulations with libEnsemble, and sensitivity analysis informed optimization.
Our next OWG meeting will be on July 23, 11am.
Timur presents his work at the workshop on Research Challenges at the Interface of Machine Learning and Uncertainty Quantification
Timur attended the workshop on Research Challenges at the interface of Machine Learning and Uncertainty Quantification at the University of Southern California. This workshop brought together researchers from several DOE labs, such as Lawrence Livermore, Sandia Livermore and Albuquerque, Los Alamos, Argonne, NREL, and from several universities, among them USC, UCLA, UC Berkeley, and Michigan. Presentations and discussions were focused around several topics, among them application of machine learning methods to forecasting and inference problems in science, probabilistic learning and UQ methods for prediction and optimization of complex systems, cross-fertilization between ML and UQ communities, and challenges in introducing scientific rigor into ad hoc ML methods used in industry.
Timur presented a talk titled "Adaptive construction of Gaussian process surrogates for Bayesian solution of inverse problems".
Next OWG meetings
Our next Optimization Working Group meetings will take place on
- June 11: Introduction to Julia
- June 25: Journal discussion
Regular OWG Meetings
We will be meeting regularly (every other Monday) at 11am to discus optimization related topics (algorithms, application problems, software, etc.). If you'd like to be added to the mailing list, please email JulianeMueller@lbl.gov. The next meetings are planned for
- April 2, 11am, 59-4022
- April 16, 11am, 59-4022
- April 30, 11am, 59-4022
OWG Meeting 03/19/18
In this meeting, we have domain scientists from Neuroscience (Roy Ben-Shalom, UCSF) and Molecular Sciences (Brandon Krull, LBNL) discuss with us their optimization application problems, the approaches they took to solve them and analyze their results.
Announcement: Optimization Working Group (OWG) Meeting 03/05/18, 11am, 59-4022
In this meeting, we will discuss a possible classification (or categorization) of optimization problems. What are the important characteristics and elements that determine into which class your difficult optimization problem falls? Are there off-the-shelf solution methods that you can simply download and apply or do you need something more specific? Let's find out!
Announcement: Optimization Working Group Meeting 02/16/18, 10am, 59-4022
Our next meeting will be fairly free format. The goal is to get to know each other, to explore (common) interests in optimization, application areas and solution methods. We'd also like to brainstorm ideas for the format and frequency of these meetings. Chime in if you're interested! Can't make it? Email ideas/topics to firstname.lastname@example.org
Edit: Meeting notes are currently available only to participants. Email email@example.com if you would like to join.
Seminar announcement: 01/26/18, 11am-12pm, 59-4101
Professor Hans Mittlemann will speak about "Optimization for the Masses - NEOS, Benchmarks and (un)expected Progress". This seminar will kick off the optimization working group meetings, to be organized monthly. All are welcome to join!
Timur joined our optimization effort at the beginning of this year. He will be working on the development of algorithms for optimization problems under uncertainty.
I presented my work on algorithms for problems that have computationally cheap objective functions and computationally expensive constraints. This work is published in the Journal of global optimization.
We are looking for a postdoc in optimization!
The Center for Computational Sciences and Engineering has an immediate opening for a postdoctoral researcher to work on the development of computational optimization algorithms for solving computationally expensive black-box optimization problems, in particular problems with uncertainty. We are looking for a candidate with experience in nonlinear optimization, including mixed-integer, derivative-free, or global optimization, who is interested in developing new solution approaches that integrate uncertainty quantification methods in optimization algorithms. The candidate will work independently as well as collaboratively in a multidisciplinary team environment that includes mathematicians, computer/computational scientists, and domain scientists. The candidate will have the opportunity to apply their developments to exciting application areas.
For more information about the optimization at LBL, please visit https://optimization.lbl.gov
For more information about this particular opening, please contact Juli Mueller at JulianeMueller (AT) lbl.gov
To apply, please go to the posting at
The posting shall remain open until the position is filled, however for full consideration, please apply by October 7, 2017.
Juliane Müller, a research scientist in CRD's Center for Computational Sciences and Engineering was profiled by INFORMS. She speaks about the inspiration behind her latest publication "SOCEMO: Surrogate Optimization of Computationally Expensive Multiobjective Problems," the experiences of women and minorities in the sciences and offers advice some advice younger researchers: Be yourself. More
https://optimization.lbl.gov goes live!