"Predicting Active Nematics Experiments with Deep Neural Networks"
Ruoshi Liu*, Mike Norton, Pengyu Hong, Seth Fraden
Active nematics are a class of far-from-equilibrium materials characterized by local orientational order of force-generating, anisotropic constitutes; they include bacterial films, animal cell cultures and synthetic systems comprised of reconstituted biomolecules. Pioneering the soft matter research, the Fraden Lab has strived to develop methods to control the behaviors of active nematics. First we need to develop mathematical or computational models to predict its movement. To achieve this goal, we use Deep Neural Networks (DNN) to study the spatiotemporal behavior of active nematics. The microtubule system is characterized by a spatially varying orientation field. Using Convolutional LSTM (ConvLSTM), we forecast the time evolution of this field using a few frames of experimental data. Our model is able to produce reasonable prediction as far as 30s into the future and we compare these synthesized orientation fields with real experimental data.
Support: SMURF (Summer MRSEC Undergrad Research Fellowship)