Collaborative Research: Elements: Scalable and Automated Learning of Active Dynamics (SALAD)

Active dynamics encompasses a wide range of collective behaviors exhibited by flocks of birds, schools of fish, layers of cells, and networks of filamentous proteins. In all these examples, out-of-equilibrium organization emerges spontaneously from interactions among active agents that consume energy from an internal reservoir or derive it from their surroundings. This project supports the development of scalable and automated software to simulate active dynamics and to infer the laws that govern it from the analysis of state-of-the-art experimental data, such as high-resolution microscopy videos. The goals of this project are to empower researchers to reliably extract hidden rules from noisy experimental observations of active dynamics, lower the barrier for analyzing large microscopy videos, reduce the time-consuming reimplementation of simulation and estimation, and promote cross-disciplinary collaborations. The dynamics of active matter, such as collections of fibroblasts or epithelial cells, is intrinsically stochastic and out of thermal equilibrium, and affected by a variety of complex processes, such as cell division. The large intrinsic fluctuations present in active matter systems hinder the efficient extraction of signals from noisy experimental data and thus it urgently demands the development of data-science-enabled tools to accelerate the analysis and improve the reproducibility of the findings. The development of the Dynamics Lab ecosystem addresses this urgent need by establishing two classes of interconnected software packages for real-space and scattering-based analysis of microscopy data. Together, these tools enable visualization and integration of physics-based simulations and statistical machine learning in both real space and Fourier space. Furthermore, to address the practical concerns of data sharing, such as size limit, this project supports the development of an efficient paradigm for data acquisition of active dynamics, where large raw data are stored offline, and small online data sets that sufficiently capture the raw data set are easily transferred and used for most research purposes. A major goal of Dynamics Lab infrastructure is to achieve sustained impacts on basic and applied sciences, ranging from biophysics to biomimetic materials. This award by the NSF Office of Advanced Cyberinfrastructure is jointly supported by the Division of Materials Research and the Division of Mathematical Sciences. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.