Using Genetic Programming with Prior Formula Knowledge to Solve Symbolic Regression Problem

Algorithm 1

TrainDBN: training DBN to generate the PFKB.

Input: , , , (, are training data; , are testing data)

Output: PFKB

() Initial(DL, opts) // initial the structure of DL and the parameters opts

() DL = DLsetup(DBN, , opts) // layer-wise pre-training DL

() DL = DLtrain(DBN, , opts) // build up each layer of DL to train

() (DBN, opts) // after training each layer, passing the parameters to nn

() PFKB = train(, , , opts) // fine-tune the whole deep architecture

() accuracy = test(PFKB, , ) // accuracy is the criterion of the quality of PFKM, if it is too

  small, then re-training after adjusting the model architecture or parameters

() return PFKB