M. J. D. Powell
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
M. J. D. Powell has a total of 1 co-authors. Here are the top co-authors by number of shared publications:
R. K. Beatson
Department of Applied Mathematics and Theoretical Physics, University of CambridgeSilver Street, Cambridge CB3 9EW, England
1 shared publication
Latest Publications
- 26 January 2012
- Published by Springer Nature in Mathematical Programming
- Vol. 138 (1-2) , 475-500
Beyond symmetric Broyden for updating quadratic models in minimization without derivatives
- 26 January 2012
- Published by Springer Nature in Mathematical Programming
- Vol. 138 (1-2) , 475-500
Beyond symmetric Broyden for updating quadratic models in minimization without derivatives
- 1 June 2004
- Published by Taylor & Francis in Optimization Methods and Software
- Vol. 19 (3-4) , 399-411
On the use of quadratic models in unconstrained minimization without derivatives
Quadratic approximations to the objective function provide a way of estimating first and second derivatives in iterative algorithms for unconstrained minimization. Therefore, we address the construction of suitable quadratic models Q by interpolating values of the objective function F. On a typical iteration, the objective function is calculated at the point that minimizes the current quadratic model subject to a trust region bound, and we find that these values of F provide good information for the updating of Q, except that a few extra values are needed occasionally to avoid degeneracy. The number of interpolation points and their positions can be controlled adequately by deleting one of the current points to make room for each new one. An algorithm is described that works in this way. It is applied to an optimization calculation that has between 10 and 160 variables. The numerical results suggest that, if m = 2n + 1, then the number of evaluations of F is only of magnitude n, where m and n are the number of interpolation conditions of each model and the number of variables, respectively. This success is due to the technique that updates Q. It minimizes the Frobenius norm of the change to ∇2 Q, subject to the interpolation conditions that have been mentioned.
- 1 June 2004
- Published by Taylor & Francis in Optimization Methods and Software
- Vol. 19 (3-4) , 399-411
On the use of quadratic models in unconstrained minimization without derivatives
Quadratic approximations to the objective function provide a way of estimating first and second derivatives in iterative algorithms for unconstrained minimization. Therefore, we address the construction of suitable quadratic models Q by interpolating values of the objective function F. On a typical iteration, the objective function is calculated at the point that minimizes the current quadratic model subject to a trust region bound, and we find that these values of F provide good information for the updating of Q, except that a few extra values are needed occasionally to avoid degeneracy. The number of interpolation points and their positions can be controlled adequately by deleting one of the current points to make room for each new one. An algorithm is described that works in this way. It is applied to an optimization calculation that has between 10 and 160 variables. The numerical results suggest that, if m = 2n + 1, then the number of evaluations of F is only of magnitude n, where m and n are the number of interpolation conditions of each model and the number of variables, respectively. This success is due to the technique that updates Q. It minimizes the Frobenius norm of the change to ∇2 Q, subject to the interpolation conditions that have been mentioned.
- 1 January 1999
- Published by Springer Nature
- p. 215-232
Recent research at Cambridge on radial basis functions
- 1 January 1999
- Published by Springer Nature
- p. 215-232
Recent research at Cambridge on radial basis functions
- 1 January 1992
- Published by Oxford University Press (OUP) in IMA Journal of Numerical Analysis
- Vol. 12 (1) , 107-133
Univariate Interpolation on a Regular Finite Grid by a Multiquadric Plus a Linear Polynomial
Univariate multiquadric interpolation to a twice continuously differentiable function on a regular infinite grid enjoys second order convergence and some excellent localization properties, but numerical calculations suggest that, if the grid is finite, then usually the convergence rate deteriorates to first order near the grid boundaries, ibis conjecture is proved. It is also shown that one can recover superlinear convergence by adding a linear polynomial term to the multiquadric approximation. Making such additions is a standard technique, but we find that the usual way of choosing the polynomial fails to provide superlinear convergence m general. Therefore some new procedures are given that pick a suitable polynomial automatically. Thus it is not unusual to reduce the maximum error of the interpolation by a factor of 10 3 . Further, it is straightforward to include one of the new procedures in multiquadric interpolation to functions of several variables when the data points are in general position.
- 1 January 1992
- Published by Oxford University Press (OUP) in IMA Journal of Numerical Analysis
- Vol. 12 (1) , 107-133
Univariate Interpolation on a Regular Finite Grid by a Multiquadric Plus a Linear Polynomial
Univariate multiquadric interpolation to a twice continuously differentiable function on a regular infinite grid enjoys second order convergence and some excellent localization properties, but numerical calculations suggest that, if the grid is finite, then usually the convergence rate deteriorates to first order near the grid boundaries, ibis conjecture is proved. It is also shown that one can recover superlinear convergence by adding a linear polynomial term to the multiquadric approximation. Making such additions is a standard technique, but we find that the usual way of choosing the polynomial fails to provide superlinear convergence m general. Therefore some new procedures are given that pick a suitable polynomial automatically. Thus it is not unusual to reduce the maximum error of the interpolation by a factor of 10 3 . Further, it is straightforward to include one of the new procedures in multiquadric interpolation to functions of several variables when the data points are in general position.