Analysis of neural networks constructed from fuzzy operations

OData support
Supervisor:
Dr. Kóczy László Tamás
Department of Telecommunications and Media Informatics

Artificial neural networks have developed rapidly in theory and applied research since 1980s.

In order to get a better model, the researchers have done a lot of work on network structure, algorithms, neurons, etc.

Fuzzy flip-flop can provide nonlinear features, which makes it suitable for an activation function to bring fuzzy features into the network. Fuzzy operations based ANN are not very fast in the convergence. However, they are robust in modeling and they eliminate the problem of overfitting. According the property of fuzzy flip-flop, fuzzy flip-flop (F3) can be used to construct the activation function of neural network.

In this study, a network is built with certain common activation functions and fuzzy flip-flop based on Trigonometric fuzzy operation. After training the network using LM, compare the result of model with different activation functions and do some analysis. Then, apply the model to reduce the influence of interference current to objective parameter of the piezo-resistive pressure sensor.

Downloads

Please sign in to download the files of this thesis.