Scientific Machine Learning and Quantum Utility: A Near Future Perspective

Table of Contents

Seminars-2-2024 - This article is part of a series.

Part 1: The Dawn of an Immersive Internet: XR, Generative AI and the Road to 6G

Part 2: AGI Chips - The Next Frontier

Part 3: Generalist vs Specialist Language Models

Part 4: From Intelligent Surfaces to Noise-Driven Communication: Innovative Technologies for 6G and Beyond

Part 5: An Overview of Evolutionary Multi-Objective Optimization

Part 6: Packet Trimming at the Edge for Low Latency in 6G Environments

Part 7: This Article

Part 8: Cloud Storage Systems: Latency Characterization and Extensions

Part 9: Let the Plants do the Talking: Climate-Smart Agriculture by the Messages Received from Plants and Soil

Part 10: More is different: How the Science of Complex Systems can inspire Future Autonomous Networks

Abstract #

In 2021, Gartner Inc., a global leader in information technology research and advisory, identified Physics-Informed Machine Learning (PIML) as one of the most promising AI technologies for the next decade. PIML seamlessly integrates data with mathematical models, even in complex, uncertain, and high-dimensional contexts, utilizing neural networks or traditional regression techniques in a meshless and straightforward manner. This talk traces the evolution of PIML within IBM Research, beginning with our early experiments using Physics-Informed Neural Networks (PINNs) to address seismic wavefield problems. We will discuss both the strengths and limitations of PINNs and the challenges of applying them to real-world engineering and environmental issues. Next, we explore the development of Reduced Order Models (ROMs) for creating low-dimensional representations of high-fidelity systems, with applications to ocean and atmospheric circulation, pollution dispersion, and extreme sea surface height modeling. Finally, we will present a forward-looking perspective on extending the Non-Intrusive Operator Inference (OpInf) method into a hybrid quantum-classical algorithm. By leveraging its simple mathematical structure based on regularized ridge regression, we aim to harness quantum hardware speed-up for future advancements.

Bio of Alberto Nogueira #

Alberto Costa graduated in Mechanical Engineering from the State University of Campinas (1994). His academic excellence awarded him a scholarship in the exchange program for Engineering students at the Institut National Des Sciences Appliquées de Lyon - France - from July 1992 to November 1993. He holds a Master of Science and a Ph.D. in Mechanical Engineering from the Faculty of Mechanical Engineering, State University of Campinas (1998 and 2002, respectively), in the following subjects: (1) High Order Spectral/hp Finite Element Method for 3D unstructured meshes applied to non-linear Elasticity, (2) Algebraic Multigrid schemes and (3) scientific software development. He also concluded two post-doctorates in Computational Fluid Dynamics (CFD), one at the Faculty of Civil Engineering, State University of Campinas (2004), and the other at the Brazilian Aerospace Institute (IAE), Department of Aerospace Science and Technology - DCTA (2006). For nearly a decade (2004-2013), Alberto led Thorus Scisoft Technology company as its CEO and accumulated significant industrial and innovation experience as a technology entrepreneur. His academic and scientific skills cover many branches of Numerical Analysis, Computational Mechanics, and Scientific Machine Learning (SciML) such as High-order Discontinuous Galerkin Method (DGFEM), Spectral Element Method (SEM), Finite Element Method (FEM), Finite Volume Method (FVM), compressible and incompressible flows, turbulence modeling, chaotic dynamical systems, Physics-Informed Neural Networks (PINNs), Neural Operators and Reduced Order Models (ROMs). Since May 2013, he has been a Research Scientist at the Dynamical Systems Modeling (DSM) team within IBM Research in Brazil. In early 2018, he played a pivotal role in IBM’s scientific machine learning initiative, introducing PINNs to solve direct and inverse seismic wavefield problems. Since May 2020, he has led the IBM SciML modeling team in developing state-of-the-art reduced-order models to tackle climate change. Alberto is a member of the Open Source Science (OSSci) program and the principal evangelist of the IBM-born open-source project SimulAI ( https://github.com/IBM/simulai), a scientific toolkit designed to be a high-level collection of machine learning techniques for multiscale Physics-Informed applications. In 2024, he embarked on the Quantum Utility Era pioneered by IBM and adopted Quantum Scientific Machine Learning (QSciML) as his top-priority learning topic. One of his short to medium-term career goals is to “quantize” SimulAI’s SciML features.