About: Full Covariance Gaussian Mixture Models Evaluation on GPU

Facets (new session)
Description
Metadata
Settings
- owl:sameAs
- Inference Rule:

About: Full Covariance Gaussian Mixture Models Evaluation on GPU Goto Sponge NotDistinct Permalink

An Entity of Type : http://linked.opendata.cz/ontology/domain/vavai/Vysledek, within Data Space : linked.opendata.cz associated with source document(s)

Attributes	Values
rdf:type	skos:Concept http://linked.opendata.cz/ontology/domain/vavai/Vysledek
Description	Gaussian mixture models (GMMs) are often used in various data processing and classification tasks to model a continuous probability density in a multi-dimensional space. In cases, where the dimension of the feature space is relatively high (e.g. in the automatic speech recognition (ASR)), GMM with a higher number of Gaussians with diagonal covariances (DC) instead of full covariances (FC) is used from the two reasons. The first reason is a~problem how to estimate robust FC matrices with a~limited training data set. The second reason is a~much higher computational cost during the GMM evaluation. The first reason was addressed in many recent publications. In contrast, this paper describes an efficient implementation on Graphic Processing Unit (GPU) of the FC-GMM evaluation, which addresses the second reason. The performance was tested on acoustic models for ASR, and it is shown that even a low-end laptop GPU is capable to evaluate a large acoustic model in a fraction of the real speech time. Three variants of the algorithm were implemented and compared on various GPUs: NVIDIA CUDA, NVIDIA OpenCL, and ATI/AMD OpenCL. Gaussian mixture models (GMMs) are often used in various data processing and classification tasks to model a continuous probability density in a multi-dimensional space. In cases, where the dimension of the feature space is relatively high (e.g. in the automatic speech recognition (ASR)), GMM with a higher number of Gaussians with diagonal covariances (DC) instead of full covariances (FC) is used from the two reasons. The first reason is a~problem how to estimate robust FC matrices with a~limited training data set. The second reason is a~much higher computational cost during the GMM evaluation. The first reason was addressed in many recent publications. In contrast, this paper describes an efficient implementation on Graphic Processing Unit (GPU) of the FC-GMM evaluation, which addresses the second reason. The performance was tested on acoustic models for ASR, and it is shown that even a low-end laptop GPU is capable to evaluate a large acoustic model in a fraction of the real speech time. Three variants of the algorithm were implemented and compared on various GPUs: NVIDIA CUDA, NVIDIA OpenCL, and ATI/AMD OpenCL. (en)
Title	Full Covariance Gaussian Mixture Models Evaluation on GPU Full Covariance Gaussian Mixture Models Evaluation on GPU (en)
skos:prefLabel	Full Covariance Gaussian Mixture Models Evaluation on GPU Full Covariance Gaussian Mixture Models Evaluation on GPU (en)
skos:notation	RIV/49777513:23520/12:43916101!RIV13-MSM-23520___
http://linked.open...avai/riv/aktivita	P
http://linked.open...avai/riv/aktivity	P(ED1.1.00/02.0090), P(TE01020197)
http://linked.open...vai/riv/dodaniDat	2013
http://linked.open...aciTvurceVysledku	Vaněk, Jan Psutka, Josef Psutka jr., Josef Trmal, Jan
http://linked.open.../riv/druhVysledku	D - Článek ve sborníku
http://linked.open...iv/duvernostUdaju	S - Úplné a pravdivé údaje nepodléhající ochraně podle zvláštních právních předpisů
http://linked.open...titaPredkladatele	Západočeská univerzita v Plzni / Fakulta aplikovaných věd
http://linked.open...dnocenehoVysledku	137351
http://linked.open...ai/riv/idVysledku	RIV/49777513:23520/12:43916101
http://linked.open...riv/jazykVysledku	eng - angličtina
http://linked.open.../riv/klicovaSlova	OpenCL; CUDA; GPU; Automatic Speech Recognition; Full Covariance; Gaussian Mixture Models (en)
http://linked.open.../riv/klicoveSlovo	OpenCL Gaussian Mixture Models CUDA GPU Automatic Speech Recognition Full Covariance
http://linked.open...ontrolniKodProRIV	[66D6D7D333B2]
http://linked.open...v/mistoKonaniAkce	Vietnam, Ho Chi Minh City
http://linked.open...i/riv/mistoVydani	345 E 47TH ST, NEW YORK, NY 10017
http://linked.open...i/riv/nazevZdroje	IEEE International Symposium on Signal Processing and Information Technology
http://linked.open...in/vavai/riv/obor	JD
http://linked.open...ichTvurcuVysledku	4 (xsd:int)
http://linked.open...cetTvurcuVysledku	4 (xsd:int)
http://linked.open...vavai/riv/projekt	Centre for Applied Cybernetics 3 NTIS - New Technologies for Information Society
http://linked.open...UplatneniVysledku	2012
http://linked.open...iv/tvurceVysledku	Psutka jr., Josef Psutka, Josef Vaněk, Jan Trmal, Jan
http://linked.open...vavai/riv/typAkce	WRD - Světová
http://linked.open.../riv/zahajeniAkce	2012-12-12 (xsd:date)
number of pages	5 (xsd:int)
http://purl.org/ne...btex#hasPublisher	Institute of Electrical and Electronics Engineers ( IEEE )
https://schema.org/isbn	978-1-4673-5604-6
http://localhost/t...ganizacniJednotka	23520

Faceted Search & Find service v1.16.118 as of Jun 21 2024

Alternative Linked Data Documents: ODE Content Formats:

RDF

ODATA

Microdata

About

OpenLink Virtuoso version 07.20.3240 as of Jun 21 2024, on Linux (x86_64-pc-linux-gnu), Single-Server Edition (126 GB total memory, 58 GB memory in use)
Data on this page belongs to its respective rights holders.
Virtuoso Faceted Browser Copyright © 2009-2024 OpenLink Software