Answers are grounded in this paper and cite their sources.
We derive two-point step sizes for the steepest-descent method by approximating the secant equation. At the cost of storage of an extra iterate and gradient, these algorithms achieve better performance and cheaper computation than the classical steepest-descent method. We indicate a convergence analysis of the method in the two-dimensional quadratic case. The behaviour is highly remarkable and the analysis entirely nonstandard.
No takes yet. Share an insight, caveat, or question.
Barzilai et al. (1988) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: