Understanding Lookahead Dynamics Through Laplace Transform

  • 2025-06-16 18:20:40
  • Aniket Sanyal, Tatjana Chavdarova
  • 0

Abstract

We introduce a frequency-domain framework for convergence analysis ofhyperparameters in game optimization, leveraging High-Resolution DifferentialEquations (HRDEs) and Laplace transforms. Focusing on the Lookaheadalgorithm--characterized by gradient steps $k$ and averaging coefficient$\alpha$--we transform the discrete-time oscillatory dynamics of bilinear gamesinto the frequency domain to derive precise convergence criteria. Ourhigher-precision $O(\gamma^2)$-HRDE models yield tighter criteria, while ourfirst-order $O(\gamma)$-HRDE models offer practical guidance by prioritizingactionable hyperparameter tuning over complex closed-form solutions. Empiricalvalidation in discrete-time settings demonstrates the effectiveness of ourapproach, which may further extend to locally linear operators, offering ascalable framework for selecting hyperparameters for learning in games.

 

Quick Read (beta)

loading the full paper ...