On the block-structured distance to non-surjectivity of sublinear mappings

• 成果类型: Article
• 署名作者: Peña, J
• 署名单位: Carnegie Mellon University
• 刊物名称: MATHEMATICAL PROGRAMMING
• ISSN/ISSBN: 0025-5610
• DOI: 10.1007/s10107-004-0514-y
• 发表日期: 2005
• 页码: 561-573
• 关键词: conic linear-system; Complexity theory; convex processes; perturbations; componentwise; optimization; algorithm; PROGRAMS; matrices

摘要

We show that the block-structured distance to non-surjectivity of a set-valued sublinear mapping equals the reciprocal of a suitable block-structured norm of its inverse. This gives a natural generalization of the classical Eckart and Young identity for square invertible matrices.