Approximating the total variation distance between spin systems
This addresses a computational challenge in statistical physics and machine learning for researchers studying graphical models, but it appears incremental as it builds on existing sampling and counting techniques.
The paper tackles the problem of approximating the total variation distance between Gibbs distributions from spin systems, proposing a reduction to sampling and counting methods, with applications to models like the hardcore and Ising models.
Spin systems form an important class of undirected graphical models. For two Gibbs distributions $μ$ and $ν$ induced by two spin systems on the same graph $G = (V, E)$, we study the problem of approximating the total variation distance $d_{TV}(μ,ν)$ with an $ε$-relative error. We propose a new reduction that connects the problem of approximating the TV-distance to sampling and approximate counting. Our applications include the hardcore model and the antiferromagnetic Ising model in the uniqueness regime, the ferromagnetic Ising model, and the general Ising model satisfying the spectral condition. Additionally, we explore the computational complexity of approximating the total variation distance $d_{TV}(μ_S,ν_S)$ between two marginal distributions on an arbitrary subset $S \subseteq V$. We prove that this problem remains hard even when both $μ$ and $ν$ admit polynomial-time sampling and approximate counting algorithms.