| 摘要 |
The neural network module comprises an input layer (40), a rule base layer (46), fuzzy layers (44,48) and an output layer (56). Within each layer, nodes are provided for input/output of data, such as input nodes (42) and rule layer nodes (50). The fuzzy input layer (44) converts inputs into membership degrees to which the input values belong. The rule based layer nodes (50) decide how the input would affect the output and is activated if the input exceeds a minimum activation threshold. The neural network also has an adaptation component that aggregates selected two or mote nodes within the rule base layer depending on the input data, and to increase the minimum activation threshold of the nodes that were not selected for aggregation. This makes the non-selected nodes less "sensitive" and will prevent future misclassification.
|
