zbMATH — the first resource for mathematics

Periodicity in optimal hierarchical checkpointing schemes for adjoint computations. (English) Zbl 1381.65024
The authors show new work on algorithms for hierarchical adjoint computations. These are models used in numerical analysis for finding a gradient of a function during optimization. The authors review prior work in the space of checkpointing models and produce two new methods that are effective in the special cases of storage-limited but with no read/write cost and storage-unlimited, but with a time cost for data access. The examples are shown with pseudocode and empirical performance results are given.
65D25 Numerical differentiation
65Y20 Complexity and performance of numerical algorithms
68W40 Analysis of algorithms
49N45 Inverse problems in optimal control
90C30 Nonlinear programming
Full Text: DOI
[1] DOI: 10.1080/10556789208805505 · doi:10.1080/10556789208805505
[2] DOI: 10.1145/347837.347846 · Zbl 1137.65330 · doi:10.1145/347837.347846
[3] DOI: 10.1080/10556780410001684158 · Zbl 1086.49024 · doi:10.1080/10556780410001684158
[4] Phelps R.R., Convex Functions, Monotone Operators and Differentiability, Vol. 1364 (1993)
[5] DOI: 10.1137/080718036 · Zbl 1194.65084 · doi:10.1137/080718036
[6] DOI: 10.1137/080742439 · Zbl 1214.65038 · doi:10.1137/080742439
[7] DOI: 10.1137/080727890 · Zbl 1196.65050 · doi:10.1137/080727890
This reference list is based on information provided by the publisher or from digital mathematics libraries. Its items are heuristically matched to zbMATH identifiers and may contain data conversion errors. It attempts to reflect the references listed in the original paper as accurately as possible without claiming the completeness or perfect precision of the matching.