How I learned to stop worrying and love StopGrads: Stationarity, Convergence, and a case study on Flow Map Learning
A stopgrad regression principle is introduced, which identifies a general template for stopgrad objectives with a closed-form characterization of stationary points and their uniqueness, unifying stopgrad objectives for flow maps, reinforcement learning, and diffusion samplers.
Max W. Shen, Mark Goldstein, Zi-Chu Wang et al.
· 0 citations