
Neural network representations of multiphase Equations of State - Scientific Reports
Equations of State model relations between thermodynamic variables and are ubiquitous in scientific modelling, appearing in modern day applications ranging from Astrophysics to Climate Science. The three desired properties of a general Equation of State model are adherence to the Laws of...

This is a neural network that models the state space.
That means, if it has another neural network as it's input, it understands the dynamics of the input.
State spaces are notoriously difficult to visualize in more than 3 dimensions. For example, the state space of the simplest nerve cell is 4 dimensional, according to the Hodgkin-Huxley equation.
State maps are how we predict the evolution of things like hurricanes and tornados. We don't do a great job at that, because we can't see the whole trajectory.
But this neural network can.