A simple adaptive accelerated optimistic dual averaging method for unconstrained convex optimization in $p$-norm geometry
arXiv cs.AI · 4 Nov 2026, 04:35 AEST
A new adaptive accelerated optimistic dual averaging method has been proposed for unconstrained convex optimization. This method utilizes low-cost preconditioning based on $p$-norm geometry and offers a unified update structure across different values of p.
Key facts
- Efficient preconditioning plays an important role in large-scale first-order optimization.
- The method is line-search-free and employs a new stepsize selection technique based on an aggregated affine lower model.
- Under a generalized local smoothness condition, we derive a convergence bound that separates the effects of constant and gap-dependent curvature.
Read aloud
A new adaptive accelerated optimistic dual averaging method has been proposed for unconstrained convex optimization. This method utilizes low-cost preconditioning based on $p$-norm geometry and offers a unified update structure across different values of p. Efficient preconditioning plays an important role in large-scale first-order optimization. The method is line-search-free and employs a new stepsize selection technique based on an aggregated affine lower model. Under a generalized local smoothness condition, we derive a convergence bound that separates the effects of constant and gap-dependent curvature.
Sources in this cluster
- A simple adaptive accelerated optimistic dual averaging method for unconstrained convex optimization in $p$-norm geometry
arXiv cs.AI · 4 Nov 2026, 04:35 AEST