Atomistic computer simulations give access to increasingly accurate and predictive modelling of materials, chemical and biochemical compounds. As more complex systems become amenable to computation, the sheer amount of data produced by a simulation, as well as its intrinsic structural complexity, make it harder to extract physical
insight from modelling.
Here I will discuss two different approaches to use a computer to assist in the analysis of a simulation: using machine-learning techniques to recognize recurring structural patterns in a material, and non-linear dimensionality reduction methods to automatically coarse-grain a high-dimensional description of structural landscapes. These techniques simplify and streamline the analysis of atomistic simulations, and provide effective structural descriptors that can be used to accelerate the exploration of free-energy landscapes, making phenomena that happen on longer time scales accessible to simulation.