With recent advances in scientific data acquisition and high-performance computing, Artificial Intelligence (AI) and Machine Learning (ML) have received significant attention from applied mathematics and physics science community. From successes reported by industry, academia, and research communities, we observe that AI and ML have great potential to leverage scientific domain knowledge to support new scientific discoveries and enhance the development of physical models for complex natural and engineering systems.  

For example, deep learning supports discovery of new materials and high-energy physics from numerous computer simulations and experiments and let us learn low-dimensional manifolds underlying the acquired data in order to represent the system of interest parsimoniously and effectively. ML has offered new insights on adaptive numerical discretization schemes and numerical solvers, which are clearly distinct from traditional mathematical theories. AI also provides a new way of generalizing constitutive physics laws based on big scientific data sets. 

Despite the progress, there are still many open questions. Our current understanding is limited regarding how and why AI/ML work and why they can be predictive. AI has been shown to outperform traditional methods in many cases especially with high-dimensional, inhomogeneous data sets. However, a rigorous understanding of when AI/ML is the right approach is largely lacking: for what class of problems, underlying assumptions, available data sets, and constraints are these new methods best suited? The lack of interpretability in AI-based modeling and related scientific theories makes them insufficient for high-impact, safety-critical applications such as medical diagnoses, national security, and environmental contamination. With transparency and a clear understanding of the data-driven mechanism, desirable properties of AI should be best utilized to extend current methods in physical and engineering modeling. Handling expensive training costs and large memory requirements for ever-increasing scientific data sets becomes also important to guarantee scalable science machine learning. 

This symposium will aim to present the current state of the art and identify opportunities and gaps in AI/ML-based physics science. The symposium will focus on challenges and opportunities for increasing the scale, rigor, robustness, and reliability of physics-informed AI necessary for routine use in science and engineering applications and discuss potential researcher-AI collaborations to significantly advance diverse scientific areas and transform the way science is done. 

Submission Guidelines

We solicit extended abstracts, full papers, and poster abstracts on topics related to the above and can include recent or ongoing research, surveys, and business/use cases.

• Extended abstracts (2 to 4 pages) and full papers (up to 6 pages) will be peer-reviewed. 
• Posters can be proposed by submitting an abstract (1 to 2 pages).

All submissions should follow the AAAI format in the Author Kit and will be handled through EasyChair (https://easychair.org/conferences/?conf=sss21). The review will be double-blind to ensure academic integrity. Accepted extended abstracts and full papers shall be published in the open access proceedings site CEUR-WS.

Topics of Interest

Authors are strongly encouraged to present papers that combine and blend physical knowledge and artificial intelligence/machine learning algorithms. Topics of interest include but are not limited to the following:

• Artificial intelligence/machine learning framework that can seamlessly synthesize models, governing equations and data
• Approaches to encode scientific knowledge in machine learning method and architecture
• Architectural and algorithmic improvements for scalable physics-informed learning
• Stability and error analysis for physics-informed learning
• Software development facilitating the inclusion of physics domain knowledge in learning
• Discovery of physically interpretable laws from data
• Applications incorporating domain knowledge into machine learning

Invited Speakers

• Animashree Anandkumar, Caltech/NVIDIA
• Surya Ganguli, Stanford University
• Jan S Hesthaven, EPFL
• Nathan Kutz, University of Washington 

Committees

Organizing committee

• Jonghyun Harry Lee, University of Hawai'i at Manoa
• Eric F. Darve, Stanford University
• Peter K. Kitanidis, Stanford University
• Michael W. Mahoney, University of California, Berkeley
• Anuj Karpatne, Virginia Tech
• Matthew W. Farthing, U.S. Army Engineer Research and Development Center
• Tyler Hesser, U.S. Army Engineer Research and Development Center

Program committee

• Kevin Carlberg, Facebook
• Marta D'Elia, Sandia National Laboratories
• Ramakrishnan Kannan, Oak Ridge National Laboratory
• Paris Perdikaris, University of Pennsylvania
• Sanghyun Lee, Florida State University
• Yan Liu, University of Southern California
• Chris Rackauckaus, MIT
• Peter Sadowski, University of Hawai'i at Manoa
• Nathaniel Trask, Sandia National Laboratories
• Hongkyu Yoon, Sandia National Laboratories
• Rose Yu, University of California, San Diego

COVID-19 arrangements

Due to the COVID-19 situation, the conference will be held virtually Monday through Wednesday, March 22-24, 2021.

Contact