| 摘要 |
<p>The invention relates to a method for transforming a fuzzy logic into a neural network. According to said method, sigmoid functions are linked together to simulate membership functions in such a way that retransformation of said neural network into a fuzzy logic remains possible even after optimization of the neural network. An advantage of the method provided by the invention is that a fuzzy logic can be transformed, particularly component by component, into a neural network and that said network can then be optimized in its entirety, i.e. all components combined. The fact that the trained neural network can be retransformed makes it possible ultimately to obtain an optimized fuzzy logic. In this way, it is possible to use fuzzy system software in an advantageous manner, in particular standardized fuzzy system software, to describe the optimized fuzzy logic.</p> |
