Portland played the backdrop for this year’s SIAM Annual meeting. A bustling city in the heart of the Oregon forest and divided by the Willamette River, Portland offered great experiences in addition to the program provided by the conference. For five days at the Oregon Convention center, students, academic leaders and industry professionals congregated to present their latest research and make professional connections. The diversity in attendees was also reflected in the diversity of their research fields. Presentation topics included applications in optimization and control, numerical analysis and scientific computing and dynamics of physics, engineering, life and social sciences. For guests arriving on Sunday, the conference offered a welcome reception which gave me a chance to catch up with colleagues as I have been away from my institution for the summer. It also gave us a chance to discuss which of the many mini-symposiums we will be attending.
The first day started with an interesting mini-symposium organized by Elizabeth Wong from UCSD on applications in optimization. Some of the highlights were Thomas Bewley, also from UCSD, presenting his work on derivative-free optimization techniques and how to accelerate their convergence in non-convex domains. This was followed by Joaquim R. R. A. Martins’ presentation on solving large-scale nonlinear optimization problems in order to optimize the structures of aircraft wings. Dr. Martins
showed the audience real time animations of shapes under simulated wind conditions being morphed into the optimal shape by the proposed solvers. The session ended with a fantastic talk by trailblazer and former first female president of SIAM, Margaret Wright. Her talk, Evolution of a Mathematical Model: Optimization, Matching, and More offered an insight into her approach in trying to improve a worker area assignment problem for the New York Department of Sanitation. Traditionally, the schedule was created by one person and they sought to make the process automated. Her talk presented one of the realities we often forget as applied mathematicians. We must take into account the constraint of human perception. Although her algorithm offered an optimal solution, the solution would have upset certain employees and so they reverted to the previous way of scheduling.
On Tuesday I was fortunate enough to attend the John Von Neumann Lecture: Untangling Random Polygons and Other Things by Charles F. Van Loan at Cornell University. Dr. Van Loan presented his work on the correlation of geometric operations on a polygon to the power method process. It was a project that evolved from a last minute homework problem he assigned to his computer science students and a lesson that our next research question is just around the corner. Later that day marked the beginning of the Workshop Celebrating Diversity (WCD). The WCD is a series of mini symposiums that allows minority graduate students to present their work and interact with professionals at various levels in their career. The mini symposium was titled Nonlinear Dynamics and Complex Systems. The speakers attempted to model and gain insight into a variety of topics. The first speaker, Selenne Banuelos of CSU Channel Islands, who co-organized the symposium with Shelby Wilson of Morehouse College, offered some insight into the effects of temperature on sleeping by modeling the dynamics of the human sleep/wake cycle under thermo-regulation. Her results accurately reproduced those found in previous experimental data. She was followed by Heather A. Brooks who presented her work on a novel way to produce Turing pattern formation in order to explain the physical structure of the ventral cord of C-elegans. Finally, I was able to present my results in a talk titled Deep Neural Networks for Low-resolution Photon-limited Imaging in which I model a photon-limited process using a Poisson distribution and implement deep neural networks to recover photon limited signals from compressed noisy observations.
The next day my morning started with another mini-symposium titled Recent Advances in Optimization Modeling and Algorithms organized by Sven Leyffer of Argonne National Laboratory. The first talk was given by my advisor Roummel Marcia. He and his group proposed quasi-Newton trust region methods as an alternative to stochastic gradient descent in machine learning algorithms. The advantage is that stochastic gradient descent typically requires the tuning of parameters while the proposed algorithm does not. Other highlights of the symposium included Cynthia Phillips from Sandia National Laboratory whose presentation described a cybersecurity problem modeled by an abstract game with an attacker and defender competing for resources. The goal of the model is to compute the optimal defense against the attacker. The seminar concluded with Zichao Di also from Argonne National Laboratory presenting her work on minimizing the error associated with advanced tomographic imaging techniques and Sven Leyffer also from Argonne National Laboratory presenting his work on Derivative free techniques for optimization with integer variables. After the mini-symposium, Jeffrey Humpherys from Brigham Young University had an interesting take on how universities design their curriculum for Math majors. He uses his school’s program which emphasizes the importance of skills that jobs in industry desire. The school also keeps a tight network of alumni in order to place students in some of the industry's most competitive jobs.
Later that day was what I considered the pinnacle of my conference experience, the Workshop Celebrating Diversity luncheon. The luncheon organized by Shelby Wilson and Erica Graham from Bryn Mawr college was designed to allow graduate students to network with national laboratory scientists, professors and industry professionals. I had the great fortune to be seated in between Esmond Ng from Lawrence Berkley Laboratory and SIAM Vice President-at-Large Carol Woodward. I enjoyed tales about their professional journeys and clung onto every piece of insight they had about my career and their descriptions of working in a laboratory setting. I cannot thank the organizers enough for inviting me to the luncheon it is something that I will cherish long after.
The rest of that day was devoted to two symposiums. First I attended two talks from the second part of the SIAM Workshop Celebrating Diversity. The symposium titled Analysis, Algorithms and Simulations for the Study of Physical Phenomena organized by Ricardo Cortez of Tulane University included Lisa J. Fauci also of Tulane University who presented her work on Hydrodynamic Synchronization in Models of Internally-driven Cilia. Her model attempts to capture the phenomenon of the synchronization of micro swimmers such as sperm in viscous fluid. Namdi Brandon of the University of North Carolina at Chapel Hill presented some interesting work on deferred correction methods for time dependent differential equation. I then attended a talk during the second part of the Recent Advances in Numerical Methods for Maxwell’s Equations in Complex Media. Camille Carvalho from the University of California, Merced, presented her work on Multiscale Modeling to capture near fields in plasmonic structures. Her method addresses the multiple scales inherent in electromagnetic problems in order captures phenomena ignored by standard methods.
SIAM Annual 2018 was a great experience. I was able to attend a number of interesting talks from a variety of disciplines. The quality of the conference was further enhanced by the venue and the host city. The Workshop Celebrating Diversity was the crown jewel of the experience. I hope that SIAM continues to incorporate these events as they are not only important for inclusion, but for creating opportunities for future scientists and academics. I look forward to next year’s meeting.
Pictures from the conference can be found on SIAM Facebook page.