Two-Point Step Size Gradient Methods
Jonathan Barzilai, Jonathan M. Borwein
Abstract
Jonathan Barzilai, Jonathan M. Borwein
Abstract
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.
OpenAlex reports 2924 citations for this work. Citation counts describe recorded attention and do not establish research quality.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
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.
Key concepts: Mathematics, Gradient descent, Quadratic equation, Computation, Method of steepest descent, Convergence (economics), Point (geometry), Applied mathematics