Gradient flows of the entropy for finite Markov chains (Q719491)

Let \(K\) be an irreducible and reversible Markov kernel on a finite set \(\mathcal{X}\), and consider the continuous time heat semigroup on \(\mathcal{X}\) associated with \(K\). It seems natural to ask whether the heat flow can also be identified as the gradient flow of an entropy functional with respect to some metric on the space of probability densities on \(\mathcal{X}\). The author argues that it is easily seen that the \(L^2\)-Wasserstein metric over a discrete space is not appropriate for this purpose; the metric derivative of the heat flow in the Wasserstein metric is typically infinite in a discrete setting, hence the heat flow cannot be interpret as the gradient flow of any functional in the \(L^2\)-Wasserstein metric. To address this issue, the author constructs a metric \(\mathcal{W}\) on the set of probability measures on \(\mathcal{X}\), which allows him to extend the interpretation of the heat flow as the gradient flow of the entropy to the setting of finite Markov chains. This result can be seen as a discrete counterpart of the Wasserstein gradient flow interpretation of the heat flow in \(\mathbb{R}^n\) by \textit{R. Jordan, D. Kinderlehrer} and \textit{F. Otto} [SIAM J. Math. Anal. 29, No. 1, 1--17 (1998; Zbl 0915.35120)]. Furthermore, the metric \(\mathcal{W}\) is similar to, but different from, the \(L^2\)-Wasserstein metric, and is defined via a discrete variant of the Benamou-Brenier formula.