Neural ODE’s - Theory and Application: From Residual Networks to Continuous-Depth Dynamics

This paper explores Neural Ordinary Differential Equations (Neural ODE’s) and then applies them to a real-world example of bacteria growth. We solve the population Initial Value Problem analytically using Bernoulli form and then compare it to the dynamics learned by a neural network. We demonstrate that Neural ODE’s can accurately reconstruct continuous, dynamic systems from discrete data.