A new sixth-order penalty function makes stochastic optimization over expectation-formulated generalized Stiefel manifolds equivalent to unconstrained optimization, enabling stochastic gradient methods with O(epsilon^-4) sample complexity.
Trust-region methods on riemannian manifolds
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
math.OC 1years
2024 1verdicts
CONDITIONAL 1roles
background 1polarities
background 1representative citing papers
citing papers explorer
-
Stochastic optimization over expectation-formulated generalized Stiefel manifold
A new sixth-order penalty function makes stochastic optimization over expectation-formulated generalized Stiefel manifolds equivalent to unconstrained optimization, enabling stochastic gradient methods with O(epsilon^-4) sample complexity.