Use this link to cite:
https://hdl.handle.net/2183/48316 Optimización de métodos de selección de características para dispositivos con recursos limitados
Loading...
Identifiers
Publication date
Authors
Noya Domínguez, Simón
Other responsabilities
Universidade da Coruña. Facultade de Informática
Journal Title
Bibliographic citation
Type of academic work
Academic degree
Abstract
[Resumo]: Ante o crecemento exponencial de dispositivos do Internet das Cousas nunha gran diversida de de ámbitos, xérase diariamente un volume de datos que require un procesamento eficiente debido ás limitacións computacionais deste tipo de dispositivos. Este traballo céntrase no estudo do efecto do uso da precisión reducida mixta aplicándose a métodos de selección de características, os cales son técnicas de preprocesado que buscan reducir a dimensionalidade dos datos sen perder rendemento, algo cada vez máis necesario. No estudo centrarémonos nos métodos de selección baseados en información mutua e como pode abordarse a redu ción da precisión utilizada neste cálculo, mediante a comparación dos resultados deste xeito cos obtidos polo uso de dobre precisión sobre varios conxuntos de datos, tanto reais como sintéticos.
[Abstract]: Given the exponential growth of Internet of Things (IoT) devices in a wide variety of areas, a volume of data is generated daily, which requires efficient processing due to the computa tional limitations of this type of devices. This work will focus on the study of the effect of using mixed reduced precision applied to feature selection methods, which are preprocessing steps that seek to reduce the dimensionality of the data, something increasingly necessary. In the study, we will focus on selection methods based on mutual information and how the reduction of the precision used in this calculation can be addressed, by comparing the results in this way with those obtained by using double precision on several data sets, both real and synthetic.
[Abstract]: Given the exponential growth of Internet of Things (IoT) devices in a wide variety of areas, a volume of data is generated daily, which requires efficient processing due to the computa tional limitations of this type of devices. This work will focus on the study of the effect of using mixed reduced precision applied to feature selection methods, which are preprocessing steps that seek to reduce the dimensionality of the data, something increasingly necessary. In the study, we will focus on selection methods based on mutual information and how the reduction of the precision used in this calculation can be addressed, by comparing the results in this way with those obtained by using double precision on several data sets, both real and synthetic.
Description
Editor version
Rights
Attribution-NonCommercial-NoDerivatives 4.0 International








