Interval arithmetic and automatic differentiation on GPU using OpenCL

Grzegorz Kozikowski; Bartłomiej Jacek Kubica

Interval arithmetic and automatic differentiation on GPU using OpenCL

Grzegorz Kozikowski, Bartłomiej Jacek Kubica

A&F Accounting & Finance

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

Abstract

This paper investigates efficient and powerful approach to the Gradient and the Hessian evaluation for complex functions. The idea is to apply the parallel GPU architecture and the Automatic Differentiation methods. In order to achieve better accuracy, the interval arithmetic is used. Considerations are based on sequential and parallel authors' implementation. In this solution, both the AD methods: Forward and Reverse modes are employed. Computational experiments include analysis of performance and are studied on the generated test functions with a given complexity. © 2013 Springer-Verlag.

Original language	English
Title of host publication	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)\|Lect. Notes Comput. Sci.
Place of Publication	Helsinki
Publisher	Springer Nature
Pages	489-503
Number of pages	14
Volume	7782
ISBN (Print)	9783642368028
Publication status	Published - 2013
Event	11th International Conference on Applied Parallel and Scientific Computing, PARA 2012 - Helsinki Duration: 1 Jul 2013 → …

Conference

Conference	11th International Conference on Applied Parallel and Scientific Computing, PARA 2012
City	Helsinki
Period	1/07/13 → …

Keywords

automatic differentiation
GPGPU
interval computations
OpenCL

Cite this

@inproceedings{4d0f4a00f85144c6b51d960a7faca7c6,

title = "Interval arithmetic and automatic differentiation on GPU using OpenCL",

abstract = "This paper investigates efficient and powerful approach to the Gradient and the Hessian evaluation for complex functions. The idea is to apply the parallel GPU architecture and the Automatic Differentiation methods. In order to achieve better accuracy, the interval arithmetic is used. Considerations are based on sequential and parallel authors' implementation. In this solution, both the AD methods: Forward and Reverse modes are employed. Computational experiments include analysis of performance and are studied on the generated test functions with a given complexity. {\textcopyright} 2013 Springer-Verlag.",

keywords = "automatic differentiation, GPGPU, interval computations, OpenCL",

author = "Grzegorz Kozikowski and Kubica, {Bart{\l}omiej Jacek}",

year = "2013",

language = "English",

isbn = "9783642368028",

volume = "7782",

pages = "489--503",

booktitle = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)|Lect. Notes Comput. Sci.",

publisher = "Springer Nature",

address = "United States",

note = "11th International Conference on Applied Parallel and Scientific Computing, PARA 2012 ; Conference date: 01-07-2013",

}

Kozikowski, G & Kubica, BJ 2013, Interval arithmetic and automatic differentiation on GPU using OpenCL. in Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)|Lect. Notes Comput. Sci.. vol. 7782, Springer Nature, Helsinki, pp. 489-503, 11th International Conference on Applied Parallel and Scientific Computing, PARA 2012, Helsinki, 1/07/13.

Interval arithmetic and automatic differentiation on GPU using OpenCL. / Kozikowski, Grzegorz; Kubica, Bartłomiej Jacek.
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)|Lect. Notes Comput. Sci.. Vol. 7782 Helsinki: Springer Nature, 2013. p. 489-503.

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Interval arithmetic and automatic differentiation on GPU using OpenCL

AU - Kozikowski, Grzegorz

AU - Kubica, Bartłomiej Jacek

PY - 2013

Y1 - 2013

N2 - This paper investigates efficient and powerful approach to the Gradient and the Hessian evaluation for complex functions. The idea is to apply the parallel GPU architecture and the Automatic Differentiation methods. In order to achieve better accuracy, the interval arithmetic is used. Considerations are based on sequential and parallel authors' implementation. In this solution, both the AD methods: Forward and Reverse modes are employed. Computational experiments include analysis of performance and are studied on the generated test functions with a given complexity. © 2013 Springer-Verlag.

AB - This paper investigates efficient and powerful approach to the Gradient and the Hessian evaluation for complex functions. The idea is to apply the parallel GPU architecture and the Automatic Differentiation methods. In order to achieve better accuracy, the interval arithmetic is used. Considerations are based on sequential and parallel authors' implementation. In this solution, both the AD methods: Forward and Reverse modes are employed. Computational experiments include analysis of performance and are studied on the generated test functions with a given complexity. © 2013 Springer-Verlag.

KW - automatic differentiation

KW - GPGPU

KW - interval computations

KW - OpenCL

M3 - Conference contribution

SN - 9783642368028

VL - 7782

SP - 489

EP - 503

BT - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)|Lect. Notes Comput. Sci.

PB - Springer Nature

CY - Helsinki

T2 - 11th International Conference on Applied Parallel and Scientific Computing, PARA 2012

Y2 - 1 July 2013

ER -

Interval arithmetic and automatic differentiation on GPU using OpenCL

Abstract

Conference

Keywords

Fingerprint

Cite this