Bulletin of the American Physical Society
APS March Meeting 2023
Volume 68, Number 3
Las Vegas, Nevada (March 5-10)
Virtual (March 20-22); Time Zone: Pacific Time
Session T53: Data Science, ML and Active MatterFocus
|
Hide Abstracts |
Sponsoring Units: GDS DBIO Chair: Jerome Delhommelle, University of Massachusetts, Lowell Room: Room 307 |
Thursday, March 9, 2023 11:30AM - 11:42AM |
T53.00001: Learning accurate closures of a kinetic theory of an active fluid. Suryanarayana Maddu, Scott Weady, Michael J Shelley Recent advances in experimental control of active matter have inspired many theoretical efforts for identifying relevant control inputs to steer non-equilibrium dynamics. Large classes of active matter systems of interest can be modeled with kinetic theories derived from first principles, however the high-dimensionality of the kinetic model poses a huge challenge to simulate. Reduced-order representation based on low-order moments of the kinetic model serve as an efficient means to simulate and control active fluids, but rely on closure assumptions to approximate unresolved higher-order moments. Still, the accuracy and differentiability of the closures determine the precision with which active fluids can be controlled. Here, we present a learning framework that relies on invariant representation of tensor-valued isotropic functions to learn the closure models directly from kinetic simulations. Using a combination of sparse and nonlinear regression techniques we learn a differentiable nonlinear map between the invariants of participating tensors and coefficients of the corresponding independent tensors. The learned expressions demonstrate excellent approximation power in comparison with commonly used closure models and approximate well beyond the parameter regime in which they were inferred. |
Thursday, March 9, 2023 11:42AM - 12:18PM |
T53.00002: Data-driven approaches to predict and understand the dynamics of active nematics Invited Speaker: Michael F Hagan Active matter is comprised of particles that generate forces, which leads to spectacular emergent dynamics resembling the lifelike properties of biological organisms. Yet, active materials exhibit diverse dynamical states, most of which have chaotic dynamics that do not produce work or other functions. Designing or controlling an active material to select a state corresponding to a desirable function requires accurate dynamical models. However, developing models using traditional statistical physics approaches is challenging because active materials lack the scale separation characteristic of equilibrium systems. |
Thursday, March 9, 2023 12:18PM - 12:30PM |
T53.00003: Directing assembly and encoding information in active matter via light patterns Jerome P Delhommelle, Caroline Desgranges The ability of active matter to assemble into reconfigurable nonequilibrium structures has drawn considerable interest in recent years. A unique property of active particles is the relation that exists between the steady-state particle density and the local self-propulsion velocity. This provides a novel avenue for promoting self-assembly through the spatial control of activity. In experiments, this has enabled to trigger the assembly of synthetic self-propelled particles, for which the application of a specific light pattern has allowed for the programmable, light-induced, self-assembly of active rectification devices. Here, using simulations and data science, we unravel the interplay between the properties of the active particles and the features of the light pattern, propose protocols to control smart templated assembly and motion in active matter, and identify a metric to quantify the amount of information encoded in the active fluid following the application of the light pattern. |
Thursday, March 9, 2023 12:30PM - 12:42PM |
T53.00004: Learning a compact representation of the nematic director of active nematics Phu N Tran, Zvonimir Dogic, Aparna Baskaran, Pengyu Hong, Michael F Hagan Machine learning has become an important tool for obtaining insights from large experimental datasets on active materials. Because they are intrinsically nonequilibrium and characterized by a broad spectrum of length and time scales, it has been challenging to develop accurate models for the dynamics of active materials using standard statistical physics approaches. In data-driven approaches, neural networks directly learn spatiotemporal correlations in the data to predict future dynamics. In this work, we develop a neural network to learn a compact (reduced-dimension) representation of the Q-tensor field that describes the nematic director of experimental active nematics. Prediction of the active nematics dynamics is then performed in the learned low dimensional space. The compact representation may help to reduce the computation complexity in important downstream tasks such as forecasting and controlling of active nematics. |
Thursday, March 9, 2023 12:42PM - 12:54PM |
T53.00005: Categorizing spatiotemporal dynamics of bacterial swarm fronts Alasdair Hastewell, Hannah Jeckel, Andreea-Oana Chelban, Gabriel Rodriguez-Roig, Knut Drescher, Jorn Dunkel Low-dimensional effective models have proved to be an essential tool for analyzing extensive high-dimensional complex biophysical data, enabling computationally efficient characterizations of the dynamics of living systems. Recent advances in automated experimental imaging allow recording the collective motion of bacterial swarms across Bacillus Subtilis' single-gene knockouts, whose morphology exhibit a rich breadth of macroscopic phenomenology based on their genotype. Here, we reduce the complex dynamics of the multicellular system to the time evolution of closed curves by representing the swarms by their moving boundary. The curves provide a three-dimensional spacetime surface representation of each mutant's phenomenology. We model these spacetime surfaces using a simple geometric model incorporating gauge invariances and physical constraints. Utilizing modern inference techniques for dynamical systems, we infer the parameters of this simplified model for each mutant and use the results to cluster the spatiotemporal phenotype of the swarm shape dynamics under varying genotypes. |
Thursday, March 9, 2023 12:54PM - 1:06PM |
T53.00006: Predicting spatio-temporal patterns of cells guided by time-varying guidance cues with reservoir computing Hoony Kang, Keshav Srinivasan, Michelle Girvan, Wolfgang Losert While the ability of reservoir computers to predict the temporal evolution of certain dynamical systems has been demonstrated extensively, it is limited to the portion of the phase space that the reservoir is shown during the learning process. This problem is further exacerbated when the training data is short in time and includes a stochastic component. We therefore propose a reservoir computer architecture with an additional input vector that mimics the temporal dynamics of a parameter of the dynamical system. Through this, the reservoir is able to learn the correspondence between the dynamics and its parameter value. We show that the reservoir is able to predict not only the steady state climate of the parameter regime, but also the climate of the transient behavior of the dynamical system when the parameter value is suddenly changed. Finally, the ability of the reservoir to predict the actual local dynamics is also assessed. |
Thursday, March 9, 2023 1:06PM - 1:18PM |
T53.00007: Rapid detection and classification of motile cell tracks in 3D Samuel A Matthews, Laurence G Wilson, James Walker, Victoria Hodge Tracks of motile microbes can be used to identify species, such as pathogens, with different swimming behaviours. They provide detailed information on responses to external stimuli such as chemical gradients and physical objects. Digital holographic microscopy (DHM) is a well-established, but computationally intensive method for obtaining three-dimensional cell tracks from video microscopy data. We use DHM data as ground truth libraries for a deep learning object detection network. The trained network allows a 100-fold increase in processing speed, and is suitable for implementation in real-time applications on modest computing hardware. Furthermore, we explore a range of machine learning tools for track classification and discuss potential applications in species identification and life detection. |
Thursday, March 9, 2023 1:18PM - 1:30PM |
T53.00008: Neural networks for data-driven models of cell mechanics Matthew Schmitt, Jonathan Colen, Stefano Sala, Margaret Gardel, Patrick W Oakes, Vincenzo Vitelli Mechanical behaviors of cells arise through the mechanochemical interactions of proteins which self-organize into organelles and cytoskeletal structures. However, no systematic strategy exists to identify the relevant collective variables representing protein distributions within the cell and link these to mechanical response at the cellular scale. Here we show how machine learning can link protein distributions to mechanical forces, leading to data-driven physical models without requiring knowledge at the microscale. We train a neural network to predict traction forces in cells from fluorescently-labeled proteins to establish a protein-force relation. From these networks, we extract an effective model of force as a Coulomb-like interaction between localized protein structures within the cell. Next, we construct a data-driven elastic cell model which bypasses microscopic theory and directly links proteins to continuum mechanical parameters. This procedure uncovers length scales of biologically-relevant features of the protein distribution. These models further allow us to perform high-throughput probes of potential biological perturbations to identify promising new experiments. |
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