Constraints in First-Order Optimization
- 20 July 2021
- Learning and Dynamical Systems
New preprint available
We recently developed a new class of first-order methods for smooth constrained optimization that are based on an analogy to non-smooth dynamical systems. The resulting algorithms are suitable for large-scale constrained optimization problems even when the feasible set fails to have a simple structure.
For more details see: https://arxiv.org/abs/2107.08225
constrained optimization; optimization for machine learning; non-smooth dynamical systems