The relative entropy between two Markov transition rate matrices from sample path considerations is derived. This relative entropy is interpreted as a 'level 2.5' large deviations action functional. That is, the level two large deviations action functional for empirical distributions of continuous-time Markov chains can be derived from the relative entropy using the contraction mapping principle.
All Science Journal Classification (ASJC) codes
- Information Systems
- Computer Science Applications
- Library and Information Sciences