FAIR for reproducible AI in scientific research applications

• Line Pouchard (Brookhaven National Laboratory)

Room: Training Room **Abstract**: AI is increasing in use and importance in scientific research, as scientists use AI to solve data-intensive problems, reduce computational costs, or replace time-consuming tasks in their scientific workflows. In parallel, scientific publications have asked the question: “is there a reproducibility crisis?” They have documented cases, surveyed the community, started investigating technical solutions, proposed recommendations, and generally study how the classic principles of research of reproducible scientific processes and results apply in this new context. In this talk, we discuss the challenges of reproducibility for AI applications, community definitions, and initial solution approaches. We will make the case that reproducibility and Findable, Accessible, Interoperable, and Re-usable (FAIR) principles are essential foundations for trustworthy AI, where users can understand and trust the results of the AI methods used. **Bio:** Line C. Pouchard is an internationally recognized expert with over 2 decades of experience in computational science in domains of interest to the Department of Energy and over 100 publications. She leads multi-disciplinary technical projects to create innovative approaches for scientific data discovery, high performance and data-intensive workflows, and FAIR data management and curation. Her present research focuses on provenance for workflows at scale, computational reproducibility, and text mining for big data. Prior to her current position at Brookhaven National Laboratory, she was Staff Scientist at Oak Ridge National Laboratory, and Assistant Professor at Purdue University. She has a PhD from the Graduate Center of the City University of New York, and an MS from the University of Tennessee, Knoxville.

Towards enabling digital twins capabilities for a cloud chamber

• Vanessa Lopez-Marrero (Brookhaven National Laboratory)

Room: Training Room **Abstract:** Particle-resolved direct numerical simulations (PR-DNS), which resolve not only the smallest turbulent eddies but also track the development and motion of individual particles, are an essential tool for studying aerosol-cloud-turbulence interactions. For instance, PR-DNS may complement experimental facilities designed to study key physical processes in a controlled environment and therefore serve as digital twins for such cloud chambers. In this talk we will present our ongoing work aimed at enabling the use of PR-DNS for this purpose. We will describe the physical model being used as well as our current efforts to improve performance and scalability of the numerical solver. This is joint work with: Jiaqi Yang (Emory Univ), Mohammad Atif (BNL), Kwangmin Yu (BNL) , Meifeng Lin (BNL), Tao Zhang (BNL), Lingda Li (BNL), Fan Yang (BNL), Yangang Liu (BNL), Abdullah Al Muti Sharfuddin (SBU), and Foluso Ladeinde (SBU). **Speaker Bio:** Dr. Vanessa Lopez-Marrero is a Computational Scientist at the Computational Science Initiative, Brookhaven National Laboratory (BNL). Prior to joining BNL she was a Research Staff Member in the Mathematical Sciences Department at the IBM T. J. Watson Research Center. In her early career she held postdoctoral appointments at the University of Illinois at Urbana-Champaign and at the Lawrence Berkeley National Laboratory. She holds a Ph.D. in Scientific Computing from the Computer Science Department, University of Illinois at Urbana-Champaign, with a Certificate of Specialization in Computational Science and Engineering, a B.A. in Mathematics from Rutgers University at New Brunswick, New Jersey, and a B.B.A. in Computer Information Systems from the University of Puerto Rico, Rio Piedras. Her general research interests are in computational and applied mathematics, dynamical systems, modeling and numerical simulation of complex systems, numerical solution of partial differential equations, numerical linear algebra, inverse problems, and scientific machine learning.

A review of quantum optimization techniques using quantum computers

• Paulo Castillo (SUNY-Farmingdale/Brookhaven National Laboratory)

Room: Training Room **Abstract**: With the development of quantum computers, many quantum algorithms are being developed to solve fundamental problems in science and engineering. Unlike classical computer, quantum computers provide a huge speed advantage when processing quantum algorithms. In this talk, we will discuss some of these fundamental problems including solving linear systems of equation and combinatorial optimization problems such as Quadratic Unconstrained Binary Optimization (QUBO) model. We will briefly mention the structure behind the quantum computer framework to solve this problem followed by the quantum annealing method, which is the optimization process used to find the global minimum of an objective function over a set of multiple solutions. **Bio**: Dr. Castillo is currently an Assistant Professor at SUNY – Farmingdale State College at the Electrical Engineering Department. He is also a research collaborator at the Computational Science Initiative Department at Brookhaven National Laboratory. His current research focuses on quantum approximate optimization algorithms. Dr. Castillo obtained his Ph.D. in Electrical Engineering from the City College of New York in 2014. His research was focused on optical remote sensing, infrared technology, optical instrumentation and atmospheric monitoring. He developed a novel field deployable open path system based on a tunable single distributed-feedback quantum cascade laser to measure the absolute spectra absorbance and concentration of gas molecules in a large non-invasive scan area.

When not to use machine learning: A perspective on potential and limitations

• Matthew Carbone (Brookhaven National Laboratory)

Room: Training Room **Abstract:** The unparalleled success of artificial intelligence (AI) in the technology sector has catalyzed an enormous amount of research in the scientific community. It has proven to be a powerful tool, but as with any rapidly developing field, the deluge of information can be overwhelming, confusing and sometimes misleading. This can make it easy to become lost in the same hype cycles that have historically ended in the periods of scarce funding and depleted expectations known as AI Winters. Furthermore, while the importance of innovative, high-risk research cannot be overstated, it is also imperative to understand the fundamental limits of available techniques, especially in young fields where the rules appear to be constantly rewritten and as the likelihood of application to high-stakes scenarios increases. In this perspective, we highlight the guiding principles of data-driven modeling, how these principles imbue models with almost magical predictive power, and how they also impose limitations on the scope of problems they can address. Particularly, understanding when not to use data-driven techniques, such as machine learning, is not something commonly explored, but is just as important as knowing how to apply the techniques properly. We hope that the discussion to follow provides researchers throughout the sciences with a better understanding of when said techniques are appropriate, the pitfalls to watch for, and most importantly, the confidence to leverage the power they can provide. **Bio:** Matthew R. Carbone is an assistant computational scientist in the Computational Science Initiative at Brookhaven National Laboratory. In 2021, he received his PhD in chemical physics at Columbia University, where he was a Department of Energy Computational Science Graduate Fellow. Currently, Matt works on problems at the intersection of physics/materials science, data-driven techniques, and computer science, such as surrogate modeling for x-ray absorption spectroscopy, structure determination and relevant software development. Matt can be reached by email at mcarbone@bnl.gov, and you can find his website at matthewcarbone.github.io.

Applied Mathematics at BNL

• Nathan Urban (Brookhaven National Laboratory)

Room: Training Room **Abstract:** Applied mathematics focuses on the development and practical use of mathematical methods to solve real-world problems. It intersects many fields including numerical methods for modeling and simulation, statistics, machine learning, operations research, control theory, and scientific computing. In this talk I give a brief overview of some areas of applied mathematics research at BNL, and will explore more deeply an application of quantifying uncertainties in climate prediction. **Bio:** Nathan Urban is the group leader of the Applied Mathematics group at Brookhaven National Laboratory's Computational Science Initiative (CSI). He holds a Ph.D. in condensed matter physics from Penn State, and has previously held research positions at Los Alamos National Laboratory, Princeton, and Penn State. His research interests include Bayesian inference and spatiotemporal statistics, probabilistic prediction and forecasting, multi-model / model-form / model structural uncertainty quantification, reduced order modeling, scientific machine learning and hybrid physical-data driven modeling, in-situ/streaming data analysis at scale, information fusion, decision making under uncertainty and optimal experimental design, and integrated multiscale computational frameworks for decision support.