A Minkowski space embedding to understand Markov models dynamics

To describe the behavior of Markov models as parameters are varied, I show how to embed the space of Markov models within a Minkowski space. This embedding maintains the inherent distance between different instances of the model. The coordinates of this embedding emerge from the symmetrized Kullback-Leibler divergence and are expressed in terms of thermodynamic quantities, organizing the Minkowski space into equilibrium and nonequilibrium components. With this approach, we can visualize models dynamics and gain a thermodynamic interpretation of information geometric concepts, even in far-from-equilibrium scenarios. I illustrate this approach using an analytically solvable molecular motor model.