Revealing hidden dynamics from time-series data by ODENet

Pipi Hu, Wuyue Yang, Yi Zhu, Liu Hong

Abstract: Highlights•A new type of interpretable neural network – ODENet is proposed.•ODENet can automatically reveal the hidden ODE dynamics from the time-series data.•Sparse regression helps ODENet to find out the simplest model fitting to the data.•ODENet shows a remarkable anti-noise performance and information detection accuracy.•Data-driven and physics-based models are constructed for actin kinetics.