Redes neuronales convolucionales y aplicaciones
Cargando...
URL Oficial
Acceso al documento en PDC
Fecha de publicación
2022
Autores
Directores (o tutores)
Editores
Título de la revista
ISSN de la revista
Título del volumen
Editor
Citas
Exportar
Citación
Resumen
Las redes neuronales artificiales (RNA) surgen en los años 40 con la intención de simular el funcionamiento de las neuronas del cerebro humano. Entre los modelos implementados en esa época destaca el Perceptrón, conocido como la unidad básica de las redes neuronales. A este primer periodo, le sigue otro de frustración y desprestigio debido al limitado soporte económico y computacional de la época. Pese a ello, algunos investigadores siguieron trabajando y se desarrolló el método de aprendizaje conocido como Backpropagation, actualmente el más utilizado en arquitecturas multicapa. A raíz de este y otros métodos, a principios de la década de 1980 se produjo un importante resurgimiento del interés en el campo de las redes neuronales. Se introduce el Neocognitrón, una red neuronal artificial jerárquica utilizada para el reconocimiento de caracteres manuscritos japoneses y otras tareas de reconocimiento de patrones. Es el origen de la arquitectura de las redes neuronales convolucionales (CNN). Estas redes tienen aplicaciones en el reconocimiento de vídeos, los sistemas de recomendación, la clasificación y segmentación de imágenes, el análisis de imágenes médicas, el procesamiento del lenguaje natural, y las series temporales financieras.
Artificial neural networks (ANN) emerged in the 1940s with the intention of simulating the functioning of neurons in the human brain. Among the models implemented at that time, the Perceptron, known as the basic unit of neural networks, stands out. This first period was followed by one of frustration and discredit due to the limited economic and computational support of the time. Despite this, some researchers continued working and developed the learning method known as Backpropagation, currently the most widely used in multilayer architectures. As a result of this and other methods, in the early 1980s there was a major resurgence of interest in the field of neural networks. Neocognitrón, a hierarchical artificial neural network used for Japanese handwritten character recognition, is introduced. It is the origin of the convolutional neural network architecture (CNN). These networks have applications in video recognition, recommender systems, image classification and segmentation, medical image analysis, natural language processing, and financial time series.
Artificial neural networks (ANN) emerged in the 1940s with the intention of simulating the functioning of neurons in the human brain. Among the models implemented at that time, the Perceptron, known as the basic unit of neural networks, stands out. This first period was followed by one of frustration and discredit due to the limited economic and computational support of the time. Despite this, some researchers continued working and developed the learning method known as Backpropagation, currently the most widely used in multilayer architectures. As a result of this and other methods, in the early 1980s there was a major resurgence of interest in the field of neural networks. Neocognitrón, a hierarchical artificial neural network used for Japanese handwritten character recognition, is introduced. It is the origin of the convolutional neural network architecture (CNN). These networks have applications in video recognition, recommender systems, image classification and segmentation, medical image analysis, natural language processing, and financial time series.