New paper establishes exact optimal stepsize schedules for SGD

A 14-page arXiv paper by Grimmer and Shen proves that monotonically decreasing linear stepsize schedules are the unique minimax-optimal choice for SGD in smooth convex optimization, ruling out fractal-type accelerations.

2026-11-02 ~ 2026-11-02 · 4 related posts

• New theory paper proves linear decay is the unique minimax-optimal stepsize schedule for SGD — konstmish · 2026-11-02
• New 14-page paper derives exact theory and minimax optimal stepsizes for SGD's last iterate — prof_grimmer · 2026-11-02
• Two New Optimization Papers: Optimal SGD Stepsizes and Stochastic Rayleigh-Ritz Ascent — prof_grimmer · 2026-11-02

1 near-duplicate retellings: prof_grimmer