A note on using performance and data profiles for training algorithms

Margherita Porcelli, Philippe L. Toint

Research output: Contribution to journalArticle

4 Downloads (Pure)

Abstract

This article shows how to use performance and data profile benchmarking tools to improve the performance of algorithms. We propose to achieve this goal by defining and approximately solving suitable optimization problems involving the parameters of the algorithm under consideration. Because these problems do not have derivatives and may involve integer variables, we suggest using a mixed-integer derivative-free optimizer for this task. A numerical illustration is presented (using the BFO package), which indicates that the obtained gains are potentially significant.

Original languageEnglish
Article numbera20
JournalACM Transactions on Mathematical Software
Volume45
Issue number2
DOIs
Publication statusPublished - 1 Jan 2019

Fingerprint

Training Algorithm
Derivatives
Derivative-free
Integer
Benchmarking
Optimization Problem
Derivative
Profile

Keywords

  • Algorithmic design
  • Derivative-free optimization
  • Hyper-parameters optimization
  • Trainable codes

Cite this

@article{5842b4007d5c42c4b3ffd47d9a399261,
title = "A note on using performance and data profiles for training algorithms",
abstract = "This article shows how to use performance and data profile benchmarking tools to improve the performance of algorithms. We propose to achieve this goal by defining and approximately solving suitable optimization problems involving the parameters of the algorithm under consideration. Because these problems do not have derivatives and may involve integer variables, we suggest using a mixed-integer derivative-free optimizer for this task. A numerical illustration is presented (using the BFO package), which indicates that the obtained gains are potentially significant.",
keywords = "Algorithmic design, Derivative-free optimization, Hyper-parameters optimization, Trainable codes",
author = "Margherita Porcelli and Toint, {Philippe L.}",
year = "2019",
month = "1",
day = "1",
doi = "10.1145/3310362",
language = "English",
volume = "45",
journal = "ACM Transactions on Mathematical Software",
issn = "0098-3500",
publisher = "ACM Press",
number = "2",

}

A note on using performance and data profiles for training algorithms. / Porcelli, Margherita; Toint, Philippe L.

In: ACM Transactions on Mathematical Software, Vol. 45, No. 2, a20, 01.01.2019.

Research output: Contribution to journalArticle

TY - JOUR

T1 - A note on using performance and data profiles for training algorithms

AU - Porcelli, Margherita

AU - Toint, Philippe L.

PY - 2019/1/1

Y1 - 2019/1/1

N2 - This article shows how to use performance and data profile benchmarking tools to improve the performance of algorithms. We propose to achieve this goal by defining and approximately solving suitable optimization problems involving the parameters of the algorithm under consideration. Because these problems do not have derivatives and may involve integer variables, we suggest using a mixed-integer derivative-free optimizer for this task. A numerical illustration is presented (using the BFO package), which indicates that the obtained gains are potentially significant.

AB - This article shows how to use performance and data profile benchmarking tools to improve the performance of algorithms. We propose to achieve this goal by defining and approximately solving suitable optimization problems involving the parameters of the algorithm under consideration. Because these problems do not have derivatives and may involve integer variables, we suggest using a mixed-integer derivative-free optimizer for this task. A numerical illustration is presented (using the BFO package), which indicates that the obtained gains are potentially significant.

KW - Algorithmic design

KW - Derivative-free optimization

KW - Hyper-parameters optimization

KW - Trainable codes

UR - http://www.scopus.com/inward/record.url?scp=85065609489&partnerID=8YFLogxK

U2 - 10.1145/3310362

DO - 10.1145/3310362

M3 - Article

VL - 45

JO - ACM Transactions on Mathematical Software

JF - ACM Transactions on Mathematical Software

SN - 0098-3500

IS - 2

M1 - a20

ER -