Bulletin of the American Physical Society
APS March Meeting 2022
Volume 67, Number 3
Monday–Friday, March 14–18, 2022; Chicago
Session G32: Data Science, Artificial Intelligence and Machine Learning IIIFocus Recordings Available
|
Hide Abstracts |
Sponsoring Units: GDS FIAP Chair: Thomas Meitzler, United States Army Tank Automotive Research, Development and Engineering Center Room: McCormick Place W-192B |
Tuesday, March 15, 2022 11:30AM - 12:06PM |
G32.00001: Learned numerical methods for solving partial differential equations Invited Speaker: Stephan Hoyer The numerical solution of partial differential equations (PDEs) is challenging because of the need to resolve spatiotemporal features over wide length- and timescales. Often, it is computationally intractable to resolve the finest features in the solution. The only recourse is to use approximate coarse-grained representations, which aim to accurately represent long-wavelength dynamics while properly accounting for unresolved small-scale physics. Deriving such coarse-grained equations is notoriously difficult and often ad hoc. |
Tuesday, March 15, 2022 12:06PM - 12:18PM Withdrawn |
G32.00002: Machine learning study for perovskite solar cell Cheng Peng, Feng Ke, Wendy L Mao, Thomas P Devereaux, Yu Lin, Chunjing Jia Perovskites solar cells, based on organic-inorganic lead halide-based materials (MA)PbI$_3$ (MA=CH$_3$NH$_3^+$), have shown great potential in achieving high power-conversion efficiencies and low production cost in recent years. However, this type of material usually suffers from a stability issue due to the volatility of the organic MA cation, and the toxicity and environmental mobility of ionic Pb2+ are concerns for their large-scale implementation. In our study, we use artificial intelligence and machine learning to assist the design of perovskite solar cell materials. Cutting-edge neural network based deep learning algorithms, combined with large-scale dataset obtained from first-principles numerical calculations, will greatly accelerate the design process by effectively revealing the optimized materials parameters. |
Tuesday, March 15, 2022 12:18PM - 12:30PM |
G32.00003: Learning material physics from images of battery primary particles Hongbo Zhao, Haitao D Deng, William Chueh, Richard D Braatz, Martin Bazant With the development of advanced synchrotron light sources, we have an unprecedented capability to image the spatio-dynamics of lithium-ion battery materials, allowing us to directly observe lithium (de)insertion during charge and discharge. However, data-driven analysis to extract hidden information in full images remains uncharted territory. Combining in-situ scanning transmission x-ray microscopy (STXM) images of lithium iron phosphate (LFP) nanoparticles[1], phase-field models[2], and the recently developed framework for the inverse learning of physical models [3-4], we extract the free energy and reaction kinetics of LFP, a phase-separating material, which is corroborated by theory. We also simultaneously invert the spatial heterogeneity and further validate it by multimodal imaging of the same particles. The result demonstrates the possibility of learning physical quantities that are otherwise difficult to measure and offers a new approach to imaging surface heterogeneity. |
Tuesday, March 15, 2022 12:30PM - 1:06PM |
G32.00004: Learning from data: a journey from understanding the universe to predicting human diseases Invited Speaker: Valentina Salvatelli The capability to produce, store and analyse vast amount of data had enormous impact on the progress of both physical science and industry applications. In this talk I will present some recent use-cases drawing from my experience as astrophysical researcher and as data and machine learning scientist in the healthcare industry. I will explain how the same algorithms that can be used to understand the origin of the universe can be applied to predict diseases and to support clinical trials. I will also describe how machine learning and data can be used to enhance multi-wavelength instrumentation and explain some of the scientific and industry applications opened up by this technique. I will conclude highlighting some of the challenges related to the use of big data and advanced machine learning that are common across disciplines and possible strategies to overcome them. |
Tuesday, March 15, 2022 1:06PM - 1:18PM |
G32.00005: A physics-driven study of dominant regions in soccer Indranil Ghosh, Costas J Efthimiou, Gregory DeCamillis In [1], a physics-driven kinematical method was introduced to produce an improved model for dominant regions in soccer. Contrary to other similar attempts, the model maintains the deterministic nature of the Voronoi diagram forgoing any probabilistic notions. |
Tuesday, March 15, 2022 1:18PM - 1:30PM |
G32.00006: Using Multi-Task Learning for Semantic Segmentation in Agriculture Daniel Marley, Jennifer Hobbs Combining aerial imagery with computer vision and machine learning is allowing farmers to usher in a new era of precision agriculture. Traditionally, farmers could only manually inspect the edges of a field, limiting their ability to identify issues within the field. To this end, we collect high-resolution aerial imagery and use computer vision to analyze the entire field. With this information, farmers can identify problems early and accurately to reduce costs, improve yields, and ultimately further sustainability. We analyze the images with a U-Net architecture to segment the entire field into different classes representing soil, weeds, crops, and unmanaged areas including roads, waterways, and field boundaries. There is a hierarchical relationship among the different classes, wherein weeds and crops are both forms of vegetation, and managed areas contain both soil and vegetation. We explicitly encode this structure in the model by adding segmentation tasks for the different levels of the hierarchy. In addition, we add a global classification task to predict the status of the entire field. This model achieves an IoU of 40% in the single multi-class segmentation task. The addition of the hierarchical segmentation heads and field status classification improves the IoU by 4.5%. |
Tuesday, March 15, 2022 1:30PM - 2:06PM |
G32.00007: Digital health: data science in practice Invited Speaker: Jie Ren Digital technologies have been driving a revolution in healthcare and are increasingly become center-stage in our everyday lives. Digital health technologies (DHTs), like wearable devices and mobile medical apps, have great potentials to improve our ability to accurately reveal health status and disease insights, as well as implement treatments. As DHTs often generate huge amounts of very complex data, data science plays a key role in unleashing the DHTs’ power, and a research mindset is critical to uncover novel insights from such data and drive the field forward. In this talk, I will provide a broad introduction to the digital health field, including the technology landscape, the players and stakeholders, as well as applications of data science in the development and application of digital clinical measures. Opportunities are vast in the digital health field to make positive impacts on health and medical outcomes; I will provide practical advice on how one can get involved. |
Tuesday, March 15, 2022 2:06PM - 2:18PM |
G32.00008: Lattice Strain Evolution in Na-NMC Battery From Electron Diffraction Patterns Using Deep Learning Joydeep Munshi, Yuzi Liu, Guiliang Xu, Colin L Ophus, Maria K Chan The potential risk of future lithium (Li) scarcity necessitates the search for an alternative chemistry such as the sodium ion batteries (SIBs) due to the abundance and low cost of sodium (Na). The Na analogs to the Li(Ni,Mn,Co)O2 (NMC) family of cathode materials, Na-NMC, are promising as SIB cathodes. However, cycling durability of SIB cathodes has been one of the important aspects for the performance and stability of SIBs. Lattice strain evolution during electrochemical cycling plays an important role in stability and reliability of SIB cathodes. To improve strain analysis recent advances in 4D-STEM (4D-scanning transmission electron microscope) has enabled extraction of crystallographic information from the polycrystalline samples such as the cycled Na-NMC electrode. To this end, we discuss the development of a fully automated AI/ML python-based pipeline to extract strain maps from 4D-STEM diffraction dataset. We discuss one of the targeted analysis pipelines to predict Bragg disk positions from the measured electron diffraction images of cycled NaNi0.4Mn0.4Co0.2O2 electrode. Based on the improved strain map predictions we conclude the material stability and lifetime in regard to the change in lattice parameters due to the imposed strain during the electrochemical cycling. |
Tuesday, March 15, 2022 2:18PM - 2:30PM |
G32.00009: Improved Pattern Recognition in Agricultural Applications using Self-Supervised Methods for High-Resolution Longitudinal Remote Sensing Data Jennifer Hobbs, Jing Wu, David Pichler, Daniel Marley Deep learning approaches thrive in high-data regimes, however, acquiring sufficient annotations is a major bottleneck in most applications. This is particularly extreme in remote sensing applications where data is collected at petabyte scale, but only a small fraction is annotated. While there have been recent efforts to collect large agricultural datasets, even these capture only a miniscule fraction of the data available. |
Follow Us |
Engage
Become an APS Member |
My APS
Renew Membership |
Information for |
About APSThe American Physical Society (APS) is a non-profit membership organization working to advance the knowledge of physics. |
© 2024 American Physical Society
| All rights reserved | Terms of Use
| Contact Us
Headquarters
1 Physics Ellipse, College Park, MD 20740-3844
(301) 209-3200
Editorial Office
100 Motor Pkwy, Suite 110, Hauppauge, NY 11788
(631) 591-4000
Office of Public Affairs
529 14th St NW, Suite 1050, Washington, D.C. 20045-2001
(202) 662-8700