from argparse import ArgumentParser, ArgumentDefaultsHelpFormatter from tkinter import filedialog from keras.datasets import mnist from Network.Net import WTA from Network.Tools import assign_Class, Calculate_Correct from Network.Tools import Gabor_Weight_plot, plot_Weight, WeightDist from Network.Tools import plot_NonSp, plot_AvrInp, plot_MissClass, plot_ConfMtx from brian2.units import * import matplotlib.pyplot as plt import numpy as np import yaml import os import sys def Map_to_Neurons(X_spikes, Y_label): Inp_map = np.array(assign_Class(data=X_spikes, Y=Y_label)) if np.sum(Inp_map) <= 0: print('No Spikes in Excitatory Layer!!') sys.exit(0) return Inp_map def NonSpikes(X_train, X_test): Sp_train = np.sum(X_train, axis=1) Sp_test = np.sum(X_test, axis=1) NonSp_train = [idx for idx, Zero in enumerate(Sp_train) if Zero == 0] NonSp_test = [idx for idx, Zero in enumerate(Sp_test) if Zero == 0] TrueSp_test = [idx for idx, Zero in enumerate(Sp_test) if Zero != 0] return NonSp_train, NonSp_test, TrueSp_test def TrueMiss(X_output, Y_label, NonSp): miss_arg = [idx for idx, Img_label in enumerate(X_output) if Img_label != Y_label[idx]] True_miss_arg = np.setdiff1d(miss_arg, NonSp) return miss_arg, True_miss_arg def Str2bool(Val_arg): if Val_arg == "True": return True elif Val_arg == "False": return False if __name__ == "__main__": # ==================== Argument Initialization ======================== parser = ArgumentParser(formatter_class=ArgumentDefaultsHelpFormatter) parser.add_argument("-l", "--load", default='Train', type=str, help="Selector for generating plots for training or model validation") args = vars(parser.parse_args()) addr_folder = filedialog.askdirectory() file_name = [] for file in os.listdir(addr_folder): if file.endswith('.b2'): file_name.append(file) # print(file_name[-2]) test_params = np.load(addr_folder[:-5] + '/test_params.npy') # print(test_params) # =========================== Parameters ============================== with open('Network/params.yml', 'r') as file: net = yaml.safe_load(file) file.close() init_params = { 'Gabor_filter':test_params[0], 'Norm':test_params[1] } Net_init = { 'Neurons':net['Net'][0], 'Learning_Rule':net['Net'][1], 'Nearest_Neighbor':net['Net'][2], 'Pre_Offset':net['Net'][3], 'Run_test':False, 'Monitors':False } # =========================== Model =================================== (X_train, y_train), (X_test, y_test) = mnist.load_data() Mdl = WTA(Net_setup=Net_init) Mdl.net.restore(file_name[-2][:-3], addr_folder + '/' + file_name[-2]) train_data = np.load(addr_folder[:-5] + '/Activity/Train/Temp_It_40000_40000.npy') test_data = np.load(addr_folder[:-5] + '/Activity/Test/Temp_It_40000_10000.npy') Inp_test = np.load(addr_folder[:-5] + '/Activity/Test/Input/Poisson_Count_10000.npy') Test_sum = np.array(np.sum(Inp_test, axis=1)) Inp_map = Map_to_Neurons(X_spikes=train_data, Y_label=y_train) NonSp_train, NonSp_test, TrueSp_test = NonSpikes(X_train=train_data, X_test=test_data) correct, result, Class_idx = Calculate_Correct(data=test_data, Y=y_test, Class_Map=Inp_map) miss_arg, True_miss_arg = TrueMiss(X_output=result, Y_label=y_test, NonSp=NonSp_test) # ==================== Plots of Network Behavior ====================== print('Gabor:', str(test_params[0])) print('Norm:', str(test_params[1])) WeightDist(Syn=Mdl['Syn1']) if ((Net_init['Neurons']) == 100 and (init_params['Gabor_filter'] == True)): Gabor_Weight_plot(Syn1_weight=Mdl['Syn1'].w) if ((Net_init['Neurons']) == 100 and (init_params['Gabor_filter'] == False)): plot_Weight(Syn1_weight=Mdl['Syn1'].w) plot_NonSp(Label_data=y_test[NonSp_test]) plot_AvrInp(Sp_Inp=Test_sum, Y_data=y_test, NonSp_idx=NonSp_test, Miss_idx=miss_arg, Correct_idx=Class_idx) plot_MissClass(y_arr=np.bincount(result[True_miss_arg])) plot_ConfMtx(Sp_pred=result, Label_true=y_test, TrueSp_idx=TrueSp_test) plt.show(block=True)