Publication Details

Optimal Hardware Parameters Prediction for Best Energy-to-Solution of Sparse Matrix Operations Using Machine Learning Techniques

NIKL Vojtěch, ŘÍHA Lubomír, VYSOCKÝ Ondřej and ZAPLETAL Jan. Optimal Hardware Parameters Prediction for Best Energy-to-Solution of Sparse Matrix Operations Using Machine Learning Techniques. In: INFOCOMP 2018. The Eighth International Conference on Advanced Communications and Computation. Barcelona: International Academy, Research, and Industry Association, 2018, pp. 43-48. ISBN 978-1-61208-655-2. Available from: https://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=4&ved=2ahUKEwj45dbYkMfdAhUilYsKHQe1DkwQFjADegQIBxAC&url=https%3A%2F%2Fwww.thinkmind.org%2Fdownload_full.php%3Finstance%3DINFOCOMP%2B2018&usg=AOvVaw0F5eFy3SoDGqt3wTWnO1GV
Czech title
Predikce optimálních hardwarových parametrů za účelem snížení spotřeby operací nad řídkými matice pomocí neuronových sítí
Type
conference paper
Language
english
Authors
Nikl Vojtěch, Ing. (DCSY FIT BUT)
Říha Lubomír, doc. Ing., Ph.D. (VŠB-TUO)
Vysocký Ondřej, Ing. (FIT BUT)
Zapletal Jan, Ing. (VŠB-TUO)
URL
Keywords

sparse, neural networks, energy efficiency, prediction

Abstract

Combinations of 3 hardware parameters (number of threads, core and uncore frequency) were tested for each of the 4 sparse algorithms (matrix-matrix addition, matrix-matrix multiplication, matrix-vector multiplication in IJV and CSR format) on a set of several thousands matrices for the purpose of identifying the best energy-to-solution setting for each matrix and sparse operation.

On this set of data, the possibility of optimal hardware setting prediction based on the properties of each matrix were analysed for each sparse algorithm. A calculation of Pearson correlation coefficient between the matrices' properties and optimal hardware parameters showed no direct correlation (highest 0.33 for x-y, lowest -0.25 for a-b).

A neural network with back-propagation learning was used for deeper analysis to see if matrix properties correspond to hardware settings. The input neurons represented properties of given matrix, output neurons represented optimal hardware parameters. Network properties (hidden neurons per layer, hidden neuron layers, learning coefficient and learning strategy) impact on prediction accuracy were analysed and the results showed

Published
2018
Pages
43-48
Proceedings
INFOCOMP 2018
Series
The Eighth International Conference on Advanced Communications and Computation
Conference
The Eighth International Conference on Advanced Communications and Computation, Barcelona, Spain, ES
ISBN
978-1-61208-655-2
Publisher
International Academy, Research, and Industry Association
Place
Barcelona, ES
BibTeX
@INPROCEEDINGS{FITPUB11682,
   author = "Vojt\v{e}ch Nikl and Lubom\'{i}r \v{R}\'{i}ha and Ond\v{r}ej Vysock\'{y} and Jan Zapletal",
   title = "Optimal Hardware Parameters Prediction for Best Energy-to-Solution of Sparse Matrix Operations Using Machine Learning Techniques",
   pages = "43--48",
   booktitle = "INFOCOMP 2018",
   series = "The Eighth International Conference on Advanced Communications and Computation",
   year = 2018,
   location = "Barcelona, ES",
   publisher = "International Academy, Research, and Industry Association",
   ISBN = "978-1-61208-655-2",
   language = "english",
   url = "https://www.fit.vut.cz/research/publication/11682"
}
Back to top