Bulletin of the American Physical Society
APS March Meeting 2020
Volume 65, Number 1
Monday–Friday, March 2–6, 2020; Denver, Colorado
Session B68: Machine Learning and Data in Polymer PhysicsInvited
|
Hide Abstracts |
Sponsoring Units: DPOLY Chair: Jonathan Whitmer, University of Notre Dame Room: Four Seasons 4 |
Monday, March 2, 2020 11:15AM - 11:51AM |
B68.00001: Autonomous X-ray Scattering Invited Speaker: Kevin Yager This talk will cover ongoing work to develop autonomous experimentation at a synchrotron x-ray scattering beamline. Deep learning (convolutional neural networks) is used to classify x-ray detector images, with performance improving when domain-specific data transformations are exploited ("physics-aware machine-learning"). These methods can be combining with customized data healing algorithms. To close the autonomous loop, we deploy a general-purpose algorithm that selects high-value experiments to conduct, attempting to minimize both uncertainty and experimental cost. Examples from recent autonomous experiments will be presented, including measuring nanoparticle ordering, combinatorial libraries of block copolymer materials, and realtime photo-thermal processing. |
Monday, March 2, 2020 11:51AM - 12:27PM |
B68.00002: Molecular Simulations Integrated Machine Learning Study of Bottlebrush Polymers Invited Speaker: Sanket Deshmukh Thermosensitive bottlebrush polymers (BBPs) are a type of graft polymers in which thermosensitive polymer side-chains are grafted to a polymer backbone. Most of these thermosensitive polymers with lower critical solution temperature (LCST) can undergo a coil-to-globule conformational transition with increasing temperature. This further results in a change in the overall shape of the BBPs, which is one of the most important properties needed in many biomedical applications including drug delivery, molecular actuators, etc. In this talk, I will discuss our recent coarse-grained (CG) molecular dynamics (MD) simulations study of poly(N-isopropylacrylamide) (PNIPAM; LCST= 305 K) BBPs of three different shapes: 1. Worm-like, 2. Cone-like, and 3. Dumbbell-like. The CG MD simulations were performed at 290 K (below LCST) and 320 K (above LCST) in the presence of explicit CG water for 500 ns. The analysis of simulation trajectories performed using in-house computer codes and data-driven machine-learning methods suggested that the shape of BBPs has significant impact on the conformations of side-chains as compared to other structural features (e.g. grafting density and side-chain length) and temperature. |
Monday, March 2, 2020 12:27PM - 1:03PM |
B68.00003: A Transfer Learning Framework for Improving Property Prediction, Interpretability, and Chemical Discovery from Scarce Datasets Invited Speaker: Brett Savoie Machine learning (ML) is being applied in virtually all areas of the chemical sciences to advance or complement activities that have traditionally been performed with physics-based methodologies. To sustain this progress and fulfill mounting expectations, ML must grapple with the intrinsic data scarcity of many applications. While datasets for image classification, object recognition, and some molecular properties may contain millions of samples, more typical chemical applications have access to only a few hundred to a few thousand samples. In data scarce scenarios, ML models can be severely underdetermined, exhibit limited transferability, and ultimately poor predictive power. Transfer learning addresses these data limitations with methodologies that utilize data across different domains, or data with mixed provenance and sparsity to augment and robustly train data scarce models. In this talk, I will discuss a flexible transfer learning approach to address data scarcity by using chemical latent space enrichment, whereby disparate data sources are combined in joint prediction tasks. I’ll show how this approach achieves three improvements over typical supervised learning approaches, including (i) increased property prediction accuracy from scarce data sets, (ii) increased model interpretability, and (iii) increased generative potential for use in the optimization and discovery of new chemistries. The talk will conclude with an outlook of current limitations, ongoing areas of improvement, and new applications. |
Monday, March 2, 2020 1:03PM - 1:39PM |
B68.00004: Optimization of organic molecules and macromolecules using machine learning Invited Speaker: Yaroslava Yingling The number of applications of data driven materials discovery is rapidly growing. The large amount of available materials characterization and computational data, combined with high level statistical algorithms, is proving to be extremely useful in developing complex predictive models. However, in the field of soft matter, which includes complex materials such as polymers, liquids, emulsions, colloids, and gels, there is a slower adoption of informatics strategies than in adjacent fields mainly due to complexity of underlying processes and plethora of processing components that dictates the properties. In this talk, I will discuss the application of machine learning (ML) technique for optimization of ligand functionalized nanoparticles (NPs) and biopolymers. In our approach we use a combination of high throughput molecular dynamics simulations and data available from the literature to train the ML model. We address the uncertainty associated with MD simulations in the development of the model. Using this approach, we were able to design novel nanoparticle ligands capable specific desired properties driven by the specific optimization function. Our methods can significantly speed up the search for a new organic monomers (complex ligands or polymers) design based on experimental, in silico and available literature data. |
Monday, March 2, 2020 1:39PM - 2:15PM |
B68.00005: Closed-loop, sequential learning for polymer systems Invited Speaker: Kristofer Reyes Closed-loop, sequential machine learning has garnered an increasing amount of interest over the past few years in experimental science due to its ability to efficiently explore combinatorially large spaces of experimental parameters. Techniques in this new field can help accelerate scientific exploration of such spaces through the strategic selection of experiments whose outcomes could potentially yield a high amount of information. In this talk, we explore a few applications of such closed-loop, sequential design of experiments in the study of polymer systems, ranging from the optimization of polymer emulsions for use in drug delivery, to recent work in efficient phase mapping of polymerization phenomena and even real-time optimal control in driving block-copolymer evolution. While each example differs in application and experimental objectives, we will present a unified framework for the modeling and decision-making employed in each case. We will also present our work in the development of general-purpose tools designed to lower the barriers for applying these algorithms and techniques to other problems. |
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