Second-order optimality and beyond: characterization and evaluation complexity in convexly-constrained nonlinear optimization

Journal article

High-order optimality conditions for convexly-constrained nonlinear optimization problems are analyzed. A corresponding (expensive) measure of criticality for arbitrary order is proposed and extended to define high-order ∈-approximate critical points. This new measure is then used within a conceptual trust-region algorithm to show that, if deriva- tives of the objective function up to order q ≥ 1 can be evaluated and are Lipschitz continuous, then this algorithm applied to the convexly constrained problem needs at most O(∈^−(q+1)) evaluations of f and its derivatives to compute an ∈-approximate q-th order critical point. This provides the first evaluation complexity result for critical points of arbitrary order in nonlinear optimization. An example is discussed showing that the obtained evaluation complexity bounds are essentially sharp.

Authors: Cartis, C; Gould, N; Toint, P

• Publication status: Published
• Peer review status: Peer reviewed
• Publisher: Springer
• Journal: Foundations of Computational Mathematics
• Publication date: 2017-09-01
• Acceptance date: 2017-05-29
• DOI: 10.1007/s10208-017-9363-y
• EISSN: 1615-3383
• ISSN: 1615-3375
• Keywords: complexity theory; machine learning; nonlinear optimization; high-order optimality conditions