The Alpha-Procrustes geometry at alpha=1 produces Riemannian Hessians whose eigenvalues are uniformly bounded independent of the SPD matrix condition number, provided the Euclidean Hessian satisfies uniform spectral bounds.
Advances in Neural Information Processing Systems34, 8940–8953 (2021)
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
math.OC 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Riemannian Optimization over Symmetric Positive Definite Matrices with the Alpha-Procrustes Geometry
The Alpha-Procrustes geometry at alpha=1 produces Riemannian Hessians whose eigenvalues are uniformly bounded independent of the SPD matrix condition number, provided the Euclidean Hessian satisfies uniform spectral bounds.