Fast and Interpretable Dynamics for Fisher Markets via Block-Coordinate Updates

We consider the problem of large-scale Fisher market equilibrium computation through scalable first-order optimization methods. It is well-known that equilibria can be captured using structured convex programs such as Eisenberg-Gale and Shmyrev programs. Highly performant deterministic full-gradient methods have been developed for these In this paper, we develop new block-coordinate computing equilibria, show interpretations tâtonnement-style or proportional response-style dynamics where either buyers items up one at a time. reformulate solve them proximal block coordinate descent methods, class update only small number coordinates decision variable in each iteration. Leveraging recent advances convergence analysis structures equilibrium-capturing programs, establish fast rates