| 摘要 |
The object of the inventive method is to train a neural network to determine risk functions in patients in association with a first illness relating to a predetermined illness based on the predetermined training data sets. Said predetermined data sets contain objectifiable information which can be used as a yardstick to measure the pathological condition of the patient. The neural network contains a plurality of neurons arranged in several layers, in addition to synapses connecting said neurons. In the course of the training, the structure of the neural network is simplified by tracking and eliminating the synapses which play no significant role in the evolution of the risk functions. This can be done, for example, by examining a possible correlation between the influences which two sending neurons have on the same receiving neuron, and where possible, eliminating one of the two synapses connecting to the receiving neuron.
|
