Neural network increment-type feedforward algorithm based on sample increment driving

[0022] The present invention will be further described below in conjunction with accompanying drawing:

[0023] Such as figure 1 As shown, the algorithm steps of the present invention are as follows:

[0024] Step 1. In the neural network, select any time L 0 training samples Initialize the model parameters; randomly set an m×n matrix P, m is the number of hidden layer nodes, and n is the number of input nodes. Compute Composite Matrix

[0025] Step 2, calculate the input weight and hidden layer threshold;

[0026] Step 3, and then calculate the hidden layer output matrix H 0 ;

[0027] Step 4. Calculate the output weight matrix β according to the least squares method and MP generalized inverse method 0 ; Set the parameter k=0, where k is the sequence number of samples added.

[0028] Step 5, introduce a new sample X 1 , calculate the L-th difference between the new sample and the initial training sample 0 Whether there is an increment between samples, if there is a