Existing A1, A2, A3, A4, A5 five parameters as neural network input layer, B for the output of neural network parameters (B1) here, after get B, A1, A2, A3, A4, A5 and get B again as neural network input layer, or the value of B by neural network calculation, get the new output parameter B for B2 (here), by comparison, B2 than B1 closer to the actual value, that is, through the operation, improve the prediction accuracy of neural network, what is this principle, can theoretically go ahead

CodePudding user response:

Is common BP neural network, A1 - A5) measured by experiment data, through the normalized processing, not from training data, the output parameter is a efficiency value, let me put it this way, there is a solar system, through the experiment acquisition by the intensity of solar radiation, temperature, water inlet temperature and so on five main parameters that influence working efficiency of the system, the five parameters as input layer of BP neural network, output parameters for the working efficiency of the solar system (B1), then these five parameters and just get it again as the working efficiency of the system and a BP neural network input layer, is obtained by neural network to the work efficiency of the solar system (B2), will be compared with B1, B2 obviously than B1 B2 closer to the actual value is collected through the experiment of the system efficiency, namely through such operations, the neural network prediction accuracy improved, excuse me, what is the reason, can theoretically explain