if (BatchSize == 1): BP_F = 1 else: BMBP_F = 1 ELM_F = 0 ELMT_F = 0 LDA_F = 0 LDAT_F = 0 if (BP_F == 1): # Number of epochs, step and mommentum momentum = 0 trainingAlg = paC.Training_Alg("BP", [N_epochs, Learning_Rate, momentum]) if (BMBP_F == 1): # Number of epochs, step and number of partitions trainingAlg = paC.Training_Alg("BMBP", [N_epochs, Learning_Rate, BatchSize, 1]) if (ELM_F == 1): trainingAlg = paC.Training_Alg("ELM", ["bias"]) if (ELMT_F == 1): trainingAlg = paC.Training_Alg("ELMT", [500, 0.0005, 20, "bias"]) if (LDA_F == 1): trainingAlg = paC.Training_Alg("LDA", [])
initDistrib = paC.Init_Distrib("default", ["uniform", -1, 1]) # Define initialization #initDistrib = paC.Init_Distrib("default", ["normal",0,1])# Define initialization mySLFN.set_initDistrib(initDistrib) # Set the initialization mySLFN.init_Weights() # Initialize BP_F = 0 ELM_F = 1 BMBP_F = 0 FT_F = 0 # DEFINE TRAINING ALGORITHM if (BP_F == 1): # Step and number of epochs trainingAlg = paC.Training_Alg("BP", [1000, 0.0003]) if (BMBP_F == 1): # Step and number of epochs trainingAlg = paC.Training_Alg("BMBP", [100, 0.0005]) if (ELM_F == 1): trainingAlg = paC.Training_Alg("ELM", ["bias"]) if (FT_F == 1): trainingAlg = paC.Training_Alg("FT", [100, 0.0005, "bias", "normal"]) mySLFN.set_trainigAlg(trainingAlg) # Set the trainig algorithm ################################################################# ########################## Boosting #########################
regularization = paC.Regularization("NoL2", [0.000005]) ######## TRAINING ALGORITHM ######## BP_F = 1 BMBP_F = 0 ELM_F = 0 ELMT_F = 0 LDA_F = 0 LDAT_F = 0 if (BP_F == 1): # Number of epochs, step and mommentum trainingAlg = paC.Training_Alg("BP", [50, 0.01, 0.0]) if (BMBP_F == 1): # Number of epochs, step and number of partitions trainingAlg = paC.Training_Alg("BMBP", [200, 0.01, 1]) if (ELM_F == 1): trainingAlg = paC.Training_Alg("ELM", ["no_bias"]) if (ELMT_F == 1): trainingAlg = paC.Training_Alg("ELMT", [100, 0.01, 20, "bias"]) if (LDA_F == 1): trainingAlg = paC.Training_Alg("LDA", []) if (LDAT_F == 1):
regularization = paC.Regularization("L2", [0.000005]) ######## TRAINING ALGORITHM ######## BP_F = 0 BMBP_F = 1 ELM_F = 0 ELMT_F = 0 LDA_F = 0 LDAT_F = 0 if (BP_F == 1): # Number of epochs, step and mommentum trainingAlg = paC.Training_Alg("BP", [500, 0.0003, 0.2]) if (BMBP_F == 1): # Number of epochs, step and number of partitions trainingAlg = paC.Training_Alg("BMBP", [200, 0.001, 10, 3]) if (ELM_F == 1): trainingAlg = paC.Training_Alg("ELM", ["bias"]) if (ELMT_F == 1): trainingAlg = paC.Training_Alg("ELMT", [10, 0.0005, 10, "bias"]) if (LDA_F == 1): trainingAlg = paC.Training_Alg("LDA", []) if (LDAT_F == 1):
# CENTRE INITIALIZATION initCenters = paC.Init_Centers("randomSamples") # Define initialization of centres #initCenters = paC.Init_Centers("Kmeans", [300,10,"nosplit"]) # Define initialization of centres myRBF.set_initCenters (initCenters) # Set the initialization # Set the initialization BP_F = 0 ELM_F = 1 BMBP_F = 0 FT_F = 0 # DEFINE TRAINING ALGORITHM if (BP_F == 1): # Step and number of epochs trainingAlg = paC.Training_Alg("BP",[300, 0.0003 ]) if (BMBP_F == 1): # Step and number of epochs trainingAlg = paC.Training_Alg("BMBP",[1500, 0.00008]) if (ELM_F == 1): trainingAlg = paC.Training_Alg("ELM",["bias"]) if (FT_F == 1): trainingAlg = paC.Training_Alg("FT",[7000, 0.00015, "bias","normal"]) # nHs = range (10, 200,5) # mySLFN.ELM_validate(nHs, n_iter = 10) myRBF.set_trainigAlg(trainingAlg) # Set the trainig algorithm