Summary: | Submitted by Automa??o e Estat?stica (sst@bczm.ufrn.br) on 2017-04-03T19:47:14Z
No. of bitstreams: 1
AndreQuintilianoBezerraSilva_DISSERT.pdf: 6705778 bytes, checksum: 90ff28923ed0270632f9a0851890a75a (MD5) === Approved for entry into archive by Arlan Eloi Leite Silva (eloihistoriador@yahoo.com.br) on 2017-04-05T21:51:32Z (GMT) No. of bitstreams: 1
AndreQuintilianoBezerraSilva_DISSERT.pdf: 6705778 bytes, checksum: 90ff28923ed0270632f9a0851890a75a (MD5) === Made available in DSpace on 2017-04-05T21:51:32Z (GMT). No. of bitstreams: 1
AndreQuintilianoBezerraSilva_DISSERT.pdf: 6705778 bytes, checksum: 90ff28923ed0270632f9a0851890a75a (MD5)
Previous issue date: 2017-02-14 === Um dos principais objetivos da intelig?ncia artificial ? a cria??o de agentes inspirados
na intelig?ncia humana. Isso vem sendo pesquisado utilizando v?rias abordagens, e
entre as mais promissoras para o aprendizado de m?quinas est?o os sistemas simb?licos
baseados na l?gica e as redes neurais artificiais. At? a ?ltima d?cada, ambas as abordagens
progrediam de forma independente, mas os progressos obtidos em ambas as ?reas
fizeram com que os pesquisadores come?assem a investigar maneiras de integrar as duas
t?cnicas. Diversos modelos que proporcionam a integra??o h?brida ou integrada desses
m?todos inteligentes surgiram na d?cada de 90 e continuam sendo utilizadas e melhoradas
at? hoje. Esse trabalho tem como objetivo principal a implementa??o e uso do algoritmo de convers?o
neuro-simb?lica do sistema h?brido Knowledge-Based Artificial Neural Networks
(KBANN). O sistema possui a capacidade de mapear um dom?nio te?rico espec?fico de
regras (se-ent?o) em uma rede neural e refinar a rede utilizando t?cnicas de aprendizado.
Al?m disso, como o algoritmo criado por Towell et al. (1990) n?o possui a capacidade
de adquirir novos conhecimentos sem distorcer o que j? foi aprendido, utilizou-se o algoritmo
TopGen (Optiz e Shavlik, 1995) para adicionar tal capacidade a rede. O trabalho
utilizou um jogo de tabuleiro para realizar experimentos devido a quantidade e o conhecimento
existente sobre as regras do jogo. O sistema implementado obteve resultados
interessantes, mesmo com a pertuba??o do dom?nio inicial de regras (com a exclus?o
parcial), obtendo uma taxa de acerto pr?xima a 100%. Portanto, a partir dos resultados
obtidos foi poss?vel concluir que o sistema h?brido ? capaz de se sobrepor a situa??es
adversas a qual foi submetido nessa pesquisa. === One of the main goals of artificial intelligence is the creation of agents with humanlike
intelligence. This has been researched using various approaches, and among the most
prominent for machine learning are logic-based symbolic systems and artificial neural
networks. Until the last decade, both approaches have progressed independently, but
progress in both areas has led researchers to investigate ways to integrate both approaches.
Several models that provide hybrid or integrated integration of these approaches emerged
in the 1990s, and continue to be used to this day. This work has as main objective the implementation and use of the Neural-Symbolic
conversion algorithm of Knowledge-Based Artificial Neural Networks (KBANN), the system
has the ability to map a specific theoretical domain of rules (if-then) into a neural
network, and refine the network using learning techniques. In addition, since the
algorithm created by (Towell et al., 1990) does not have the capacity to acquire new
knowledge and introduce them to the neural network, the algorithm TopGen (Optiz and
Shavlik, 1995) will be used to add The network without losing the original knowledge acquired.
The work used a board game to conduct experiments due to the well established
rules of the game. The implemented system obtained interesting results, even with the
initial rule domain perturbation (with the exclusion of them), obtaining an accuracy rate
close to 100 %. Therefore, from the obtained results it was possible to conclude that
the hybrid systems are able to overlap to adverse situations which were carried out the
analyzes proposed in this research.
|