OBCD is a feasible block coordinate descent algorithm for nonsmooth composite optimization under orthogonality constraints that updates k rows per iteration, proves completeness of the scheme, stronger global block-k stationary points, and O(1/ε) complexity to ε-block-k stationary points.
Structured quasi-newton methods for optimization with orthogonality constraints
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
math.OC 1years
2023 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
A Block Coordinate Descent Method for Nonsmooth Composite Optimization under Orthogonality Constraints
OBCD is a feasible block coordinate descent algorithm for nonsmooth composite optimization under orthogonality constraints that updates k rows per iteration, proves completeness of the scheme, stronger global block-k stationary points, and O(1/ε) complexity to ε-block-k stationary points.