event
ISyE Student Seminar Series - Jacob Aguirre
Primary tabs
Entropy-Smooth Convex Optimization Cannot Be Accelerated
Abstract: We prove an Q(L/T) lower bound for first-order optimization of convex functions smooth relative to negative entropy on the simplex, showing mirror descent is nearly optimal and acceleration is generally impossible.
Unlike prior results, our prox-function is standard rather than pathological.
We extend the same non-acceleration result to quantum optimization over spectrahedra.
Media
Summary
Entropy-Smooth Convex Optimization Cannot Be Accelerated
Status
- Workflow status: Published
- Created by: Scott Jacobson
- Created: 10/09/2026
- Modified By: Scott Jacobson
- Modified: 10/09/2026
Categories
- No categories were selected.
Keywords
- No keywords were submitted.
User Data
Target Audience