Answers are grounded in this paper and cite their sources.
We derive upper bounds for induced matrix inverse p-norms, 1<p<∞, by combining row and column information through Hölder’s and Young’s inequalities. The bounds incorporate an S-partition scaling, an explicit norm-equivalence factor, and optimization over a scalar parameter. We prove that, whenever the classical unscaled row and column diagonal-dominance estimates are applicable, the optimized direct bound is at least as sharp as their endpoint interpolation. A nonsymmetric example shows that an interior exponent can yield a strict improvement even over separately optimized endpoint bounds. For p=2, the comparison matrix becomes symmetric, which allows the S-strict diagonal-dominance assumption to be replaced by the more general S-sparse Ostrowski–Brauer condition. This gives explicit bounds for both ∥A−1∥2 and the smallest singular value. In a 54-state whole-body pharmacokinetic model, the optimized bounds improve on separately optimized endpoint interpolation at all tested exponents and metabolic-clearance levels, with a 9.65% reduction for p=2 for the baseline parameter setting.
No takes yet. Share an insight, caveat, or question.
Cvetković et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: