Izboljšana hitra Gaussova transformacija v OpenCL : diplomsko delo

Markovič, Matevž (author), Lotrič, Uroš (mentor)

Diskretna Gaussova transformacija (ang. Discrete Gauss Transform, DGT) se pogosto pojavlja na področjih strojnega učenja, informatike, fizike ter ekonomije. Zaradi njene časovne neučinkovitosti se pri problemih večjih razsežnosti pogosto posega po aproksimativnih metodah, ki nam omogočajo hiter izračun njenega približka. Ena od teh metod je metoda izboljšane hitre Gaussove transformacije (ang. Improved Fast Gauss Transform, IFGT). V pričujočem diplomskem delu je predstavljena implementacija IFGT na platformi OpenCL, ki omogoča uporabo procesnih zmogljivosti grafične kartice za pohitritev celotnega izračuna. Implementacijo smo testirali z različnimi testnimi podatki tako na procesorju kot na grafični kartici. Primerjali smo jo z implementacijo DGT, ki je bila prav tako izvedena na procesorju in na grafični kartici z uporabo OpenCL. Implementacijo IFGT smo preverili s pomočjo izračuna Rényijeve entropije.

• Language: Slovenian
• Keywords: OpenCL, grafična kartica, diskretna Gaussova transformacija, izboljšana hitra Gaussova transformacija, implementacija, paralelnost, Rényijeva entropija
• Work type: Bachelor thesis/paper
• Typology: 2.11 - Undergraduate Thesis
• Organization: FRI - Faculty of Computer and Information Science
• Publisher: [M. Markovič]
• Year: 2016
• Number of pages: 81 str.
• PID: 20.500.12556/RUL-80167
• COBISS.SI-ID: 1536771523
• Publication date in RUL: 01.02.2016

Licences

• License: CC BY-SA 2.5 SI, Creative Commons Attribution-ShareAlike 2.5 Slovenia
• Link: https://creativecommons.org/licenses/by-sa/2.5/si/deed.en
• Description: You are free to reproduce and redistribute the material in any medium or format. You are free to remix, transform, and build upon the material for any purpose, even commercially. You must give appropriate credit, provide a link to the license, and indicate if changes were made. You may do so in any reasonable manner, but not in any way that suggests the licensor endorses you or your use. If you remix, transform, or build upon the material, you must distribute your contributions under the same license as the original. You may not apply legal terms or technological measures that legally restrict others from doing anything the license permits.

Secondary language

• Language: English
• Title: Improved fast Gauss transform in OpenCL

Abstract

Discrete Gauss Transfrom (DGT) commonly appears in areas such as artificial learning, informatics, physics and economy. Due to its inefficiency in terms of speed, especially when we start considering larger problems, faster approximative methods such as Improved Fast Gauss Transform (IFGT) are frequently used instead. This thesis discusses the implementation of IFGT on the OpenCL platform, which enables us to use processing capabilities of the GPU to accelerate the computation. The implementation is tested using different sets of test data both with and without the graphics card. We compare it with the implementation of DGT, which is also implemented on the CPU and on the GPU using OpenCL. We test our implementation in the computation of continuous Rényi's entropy.

• Keywords: OpenCL, graphics card, Discrete Gauss Transform, Improved Fast Gauss Transform, implementation, parallelism, Rényi entropy