A utilização de redes neurais artificiais e teoria das inteligências múltiplas no apoio ao ensino
Carregando...
Data
2017-04-13
Autores
Nascimento, Eduardo de Paula Lima [UNESP]
Título da Revista
ISSN da Revista
Título de Volume
Editor
Universidade Estadual Paulista (Unesp)
Resumo
Redes Neurais Artificiais são técnicas computacionais que apresentam um modelo matemático inspirado na estrutura neural de organismos inteligentes, e que adquirem conhecimento pela experiência. Neste trabalho apresenta-se o desenvolvimento de uma Rede Neural Artificial (RNA) capaz de reconhecer aspectos de Inteligências Múltiplas, para que seja possível direcionar alunos para conteúdos mais apropriados em ferramentas de apoio ao ensino. Foram desenvolvidas e testadas, nesta pesquisa, três Redes Neurais Artificiais (RNA), a primeira possuindo quatro entradas, correspondentes às somas de respostas dadas a perguntas cotidianas, e quatro saídas lineares utilizando a função sinal como função de ativação. A segunda rede possui doze entradas, correspondentes às respostas das perguntas cotidianas e quatro saídas contínuas utilizando a função Sigmoide para a classificação das inteligências. A terceira rede, por sua vez, possui a mesma configuração da segunda rede, porém com uma camada oculta e também foi inserido um valor à função Sigmoide, para que o treinamento apresentasse um melhor aproveitamento. Os testes foram realizados com todas as redes desenvolvidas, que eram reprogramadas, sempre que necessário, com os resultados dos testes realizados, uma vez que em alguns casos eles ainda eram inconclusivos por não conseguirem dar segurança à classificação das inteligências. Neste trabalho também são apresentados e discutidos, os resultados obtidos na utilização da Rede Neural Artificial por alunos de diversas universidades de São José do Rio Preto.
Artificial Neural Networks are computational techniques that present a mathematical model inspired by the neural structure of intelligent organisms, and can acquire knowledge through experience. This research presents the development of an Artificial Neural Network capable of recognizing aspects of Multiple Intelligences, so that it is possible to direct students to more appropriate content in tools to support education. In this research, three Artificial Neural Networks (RNA) were developed and tested. The first network has four inputs, corresponding to the sums of answers given to daily questions, and four linear outputs using the signal function as activation function. The second network has twelve inputs, corresponding to the answers of the daily questions and four continuous outputs using the Sigmoid function for the classification of intelligences. The third network, has the same configuration as the second network, but with a hidden layer and a value was also added to the Sigmoid function, so that the training could be better. The tests were performed with all the developed networks, which were reprogrammed, whenever necessary, with the results of the tests performed, since in some cases they were still inconclusive. In this paper are also presented and discussed, the results obtained in the use of Artificial Neural Network by students of universities of São José do Rio Preto.
Artificial Neural Networks are computational techniques that present a mathematical model inspired by the neural structure of intelligent organisms, and can acquire knowledge through experience. This research presents the development of an Artificial Neural Network capable of recognizing aspects of Multiple Intelligences, so that it is possible to direct students to more appropriate content in tools to support education. In this research, three Artificial Neural Networks (RNA) were developed and tested. The first network has four inputs, corresponding to the sums of answers given to daily questions, and four linear outputs using the signal function as activation function. The second network has twelve inputs, corresponding to the answers of the daily questions and four continuous outputs using the Sigmoid function for the classification of intelligences. The third network, has the same configuration as the second network, but with a hidden layer and a value was also added to the Sigmoid function, so that the training could be better. The tests were performed with all the developed networks, which were reprogrammed, whenever necessary, with the results of the tests performed, since in some cases they were still inconclusive. In this paper are also presented and discussed, the results obtained in the use of Artificial Neural Network by students of universities of São José do Rio Preto.
Descrição
Palavras-chave
Redes neurais (Computação), Inteligências múltiplas, Ensino auxiliado por computador