Physical system means that data and variables that are used to describe real world physical systems often take the form of complex physics functions with high dimensionality. Often continuous models are used to describe how these physic functions evolve over time. The complexity of these systems and their different scales makes it necessary to finely discretize both space and time. Numerical Methods are used in these systems leading to a large number of degrees of freedom. These numerical methods use known physical laws.
One example where it is possible to use deep learning models to infer physical functions within a physical system can be found in our article on Deep learning in Physics. In short deep learning models are used instead of physical functions and thus enables good simulations of a physical system with speed-ups. For example, a number of complex liquid simulations including a set of single-phase buoyancy simulations are used to demonstrate the impact of this method.
Physical System Details
We refer to the following video for more details: