Benchmarking optimization software with performance profiles
By E. Dolan, Jorge J. Moré. 2001
4,035 citations recorded by the source corpus.
Abstract.We propose performance profiles — distribution functions for a performance metric — as a tool for benchmarking and comparing optimization software. We show that performance profiles combine the best features of other tools for performance evaluation.
Subfields: Computational Theory and Mathematics, Hardware and Architecture, Numerical Analysis
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Computational Theory and Mathematics
- Lorenz T. Biegler
- Andreas Wächter
- Luís Nunes Vicente
- Christodoulos A. Floudas
- Hongchao Zhang
- William W. Hager
- Brendan O'Donoghue
- Ruth Misener
In Hardware and Architecture
- J. Demmel
- Felix Lenders
- Andreas Potschka
- Christian Kirches
- Brendan O'Donoghue
- Klaus Gärtner
- Olaf Schenk
- R. Vuduc
In Numerical Analysis
- Luís Nunes Vicente
- Lorenz T. Biegler
- Andreas Wächter
- Christodoulos A. Floudas
- Jorge Nocedal
- Hongchao Zhang
- Ruth Misener
- Ignacio E. Grossmann
Citations in the corpus
- The Scaling and Squaring Method for the Matrix Exponential Revisited (2005, background): “3 plots a performance profile [4] for the experiment.”
- Polyhedral approximation in mixed-integer convex optimization (2016, methodology): “4 Performance profile [16] (solver is the fastest within a factor of θ of the best on proportion p of instances) over the instances representable as mixed-integer second-order cone problems where we can compare with the commercial CPLEX solver.”
- Alternating direction augmented Lagrangian methods for semidefinite programming (2010, methodology): “6 by using performance profiles as proposed in [8].”
- Global convergence of a modified Fletcher–Reeves conjugate gradient method with Armijo-type line search (2006, methodology): “3, “cg-descent” stands for the CG DESCENT method with the approximate Wolfe line search proposed by Hager and Zhang [15]; “mfr” is the MFR method with the same line search as “cg-descent”; “PRP+” means the PRP+ method with the strong Wolfe line search proposed by Moré and Thuente [21].”
- Operator Splitting for a Homogeneous Embedding of the Linear Complementarity Problem (2020, methodology): “To summarize the results for each dataset we shall use Dolan-Mor´e performance profiles [19].”
- Parallel stochastic gradient algorithms for large-scale matrix completion (2013, methodology): “1 We plot performance profiles comparing NNLS and Jellyfish in Figure 3 [12].”
- On the implementation of an interior-point filter line-search algorithm for large-scale nonlinear programming (2006, methodology): “For the comparisons in the next sections we make use of the Dolan-Moré performance profiles [9].”
- COCO: a platform for comparing continuous optimizers in a black-box setting (2016, neutral): “Performance profiles are ECDFs of these runtimes relative to the respective best solver [24].”
- Trust-region quadratic methods for nonlinear systems of mixed equalities and inequalities (2009, methodology): “The reason for using the dogleg strategy instead of the Moré and Sorensen algorithm ([28]) is that in the hard case the latter strategy steps to the boundary of the trust-region even when an unconstrained minimizer of m k is safely inside.”
- Benchmarking optimization solvers for structural topology optimization (2015, methodology): “The state-ofart technique for benchmarking optimization solvers called performance profiles (Dolan and Moré 2002) is used on an illustrative and large set of topology optimization problems.”