Matrix Theory and Algorithms
External reference: https://openalex.org/T10792
- Accelerated methods improve nonlinear fractional diffusion solving
- Generic cuspidal points can be localized from eigenvalue loops
- Schur multiplier norm and its dual are expressed by minimization formulas
- Tensor nuclear norm is fully decomposable on specific subspaces
- Modified conjugate gradient method showed better numerical performance
- Sequential ridge leverage score estimates enable Nystrom KRR