Abstract:Abstract We analyze the information geometric structure of time reversibility for parametric families of irreducible transition kernels of Markov chains. We define and characterize reversible exponential families of Markov kernels, and show that irreducible and reversible Markov kernels form both a mixture family and, perhaps surprisingly, an exponential family in the set of all stochastic kernels. We propose a parametrization of the entire manifold of reversible kernels, and inspect reversible geodesics. We define information projections onto the reversible manifold, and derive closed-form expressions for the e-projection and m-projection, along with Pythagorean identities with respect to information divergence, leading to some new notion of reversiblization of Markov kernels. We show the family of edge measures pertaining to irreducible and reversible kernels also forms an exponential family among distributions over pairs. We further explore geometric properties of the reversible family, by comparing them with other remarkable families of stochastic matrices. Finally, we show that reversible kernels are, in a sense we define, the minimal exponential family generated by the m-family of symmetric kernels, and the smallest mixture family that comprises the e-family of memoryless kernels.

Spectrum of large random reversible Markov chains: two examples

Large Deviations for Markov Chains and Spectrum of Matrices

Geometric Ergodicity and the Spectral Gap of Non-Reversible Markov Chains

The spectrum of coupled random matrices

Spectral gap of nonreversible Markov chains

Singular value distribution of dense random matrices with block Markovian dependence

Irreversible local Markov chains with rapid convergence towards equilibrium

On a boundary of the central limit theorem for strictly stationary, reversible Markov chains

Reversible Markov chains: variational representations and ordering

Cutoff for mixtures of permuted Markov chains: reversible case

Maximum entropy estimation of transition probabilities of reversible Markov chains

Non-asymptotic Estimates for Markov Transition Matrices with Rigorous Error Bounds

A Tutorial on the Spectral Theory of Markov Chains

Geometric ergodicity of trans-dimensional Markov chain Monte Carlo algorithms

Spectral crossover in non-hermitian spin chains: comparison with random matrix theory

Random walk on sparse random digraphs

The Algebra of Reversible Markov Chains

Reversibility and Entropy Production of Inhomogeneous Markov Chains

A spectral dominance approach to large random matrices: part II

Information Geometry of Reversible Markov Chains

A random walk on the category of finite abelian $p$-groups