Use this link to cite:
http://hdl.handle.net/2183/34057 Clasificación distribuida on the edge
Loading...
Identifiers
Publication date
Authors
Tomé Moure, Rubén
Other responsabilities
Universidade da Coruña. Facultade de Informática
Journal Title
Bibliographic citation
Type of academic work
Academic degree
Abstract
[Resumen]: En la actualidad, el volumen de datos es enorme, dificultando el análisis e identificación de
patrones de los mismos. En este contexto, los algoritmos actuales de aprendizaje automático
no son adecuados, debido a su implementación centralizada, que nos lleva a tiempos de ejecución
demasiado elevados. Un enfoque distribuido, sin embargo, permite el tratamiento de los
datos por parte de varios nodos de forma paralela, lo que reduce las necesidades hardware. Por
todo lo mencionado, el aprendizaje distribuido está cobrando cada vez más importancia en la
actualidad. En el aprendizaje distribuido, varios nodos clasifican un subconjunto de los datos
originales, que se enviarán a un nodo final que combinará dichos resultados. Otro concepto
importante es el de aprendizaje on the edge, que consiste en procesar los datos lo más cerca
posible, evitando el envío masivo de información a estructuras externas y la problemática que
plantearía. En este trabajo, se utilizará un ordenador convencional y dispositivos Raspberry
Pi, muy interesantes ya que ofrecen un rendimiento notable a un precio muy inferior a un PC.
[Abstract]: Nowadays, the volume of data is enormous, making it difficult to analyse and identify patterns in it. In this context, current machine learning algorithms are not suitable, due to their centralised implementation, which leads to excessively high execution times. A distributed approach, however, allows data processing by several nodes in parallel, which reduces hardware needs. For all of the above, distributed learning is becoming increasingly important today. In distributed learning, several nodes classify a subset of the original data, which will be sent to a final node that will combine those results. Another important concept is the “on the edge learning”, which consists of processing the data as closely as possible, avoiding the massive sending of information to external structures and the problems that this would pose. In this work, a conventional computer and Raspberry Pi devices will be used, which are very interesting since they offer remarkable performance at a much lower price than a PC.
[Abstract]: Nowadays, the volume of data is enormous, making it difficult to analyse and identify patterns in it. In this context, current machine learning algorithms are not suitable, due to their centralised implementation, which leads to excessively high execution times. A distributed approach, however, allows data processing by several nodes in parallel, which reduces hardware needs. For all of the above, distributed learning is becoming increasingly important today. In distributed learning, several nodes classify a subset of the original data, which will be sent to a final node that will combine those results. Another important concept is the “on the edge learning”, which consists of processing the data as closely as possible, avoiding the massive sending of information to external structures and the problems that this would pose. In this work, a conventional computer and Raspberry Pi devices will be used, which are very interesting since they offer remarkable performance at a much lower price than a PC.
Description
Editor version
Rights
Atribución 3.0 España








