Most machine learning algorithms need to handle large data sets. This feature often leads to limitations on processing time and memory. The Expectation-Maximization (EM) is one of such algorithms, which is used to train one of the most commonly used parametric statistical models, the Gaussian Mixture Models (GMM). All steps of the algorithm are potentially parallelizable once they iterate over the entire data set. In this study, we propose a parallel implementation of EM for training GMM using CUDA. Experiments are performed with a UCI dataset and results show a speedup of 7 if compared to the sequential version. We have also carried out modifications to the code in order to provide better access to global memory and shared memory usage. We have achieved up to 56.4% of achieved occupancy, regardless the number of Gaussians considered in the set of experiments.
Speech interfaces have played an important role on the use of mobile devices and pervasive systems, as well as on the support for accessibility. However, Automatic Speech Recognition (ASR) systems still face several issues that limit their performance, specially due the variability and complexity of the speech communication.One of these issues concerns the presence and influence of context information. We argue that if, well handled, context information can significantly improve the speech understanding accuracy. This paper provides a formalization of context and proposes an approach to its use on the adaptation of language model in ASR systems. This proposal is applied to the development of a remote control to mobile robotic devices. Results show a decrease up to 2.67% and 3.3% on WER and WIL rates, respectively.
Direitos para esta edição cedidos à Atena Editora pelos autores. Todo o conteúdo deste livro está licenciado sob uma Licença de Atribuição Creative Commons. Atribuição-Não-Comercial-NãoDerivativos 4.0 Internacional (CC BY-NC-ND 4.0).O conteúdo dos artigos e seus dados em sua forma, correção e confiabilidade são de responsabilidade exclusiva dos autores, inclusive não representam necessariamente a posição oficial da Atena Editora. Permitido o download da obra e o compartilhamento desde que sejam atribuídos créditos aos autores, mas sem a possibilidade de alterá-la de nenhuma forma ou utilizá-la para fins comerciais.Todos os manuscritos foram previamente submetidos à avaliação cega pelos pares, membros do Conselho Editorial desta Editora, tendo sido aprovados para a publicação.A Atena Editora é comprometida em garantir a integridade editorial em todas as etapas do processo de publicação. Situações suspeitas de má conduta científica serão investigadas sob o mais alto padrão de rigor acadêmico e ético.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.