NarmaFile, OscillatorFile = 'Narma10.h5', 'ReservoirResponse.h5'
# Split between training and testing datasets
training, testing = [0, 1], [2, 3]
# Create Narma10 samples and the Multiplexed input data
if not os.path.isfile(os.getcwd() + '/' + NarmaFile):
Narma10_Input.CreateSamples(NarmaFile, len(training) + len(testing))
if os.path.isfile(os.getcwd() + '/' + OscillatorFile):
os.remove(os.getcwd() + '/' + OscillatorFile)
# Pass the Multiplexed input through the nonlinear node
if not os.path.isfile(os.getcwd() + '/' + OscillatorFile):
Driver, Virtual_Node_time, Sim, Sim_time = Oscillator_Simulation(
NarmaFile, OscillatorFile)
Vis.Reservoir_Response(Driver, Virtual_Node_time, Sim, Sim_time,
'Reservoir_Response.png')
# Train
Inputs, Ybar_train, Ytrain, NRMSE_train, Ybar_test, Ytest, NRMSE_test1 = TrainData(
NarmaFile, OscillatorFile, training, testing)
Vis.Trained_Datasets(Inputs, Ybar_train, Ytrain, NRMSE_train, training,
r'$NRMSE_{train} = %.3f$' % NRMSE_train,
'Train_Dataset.png')
Vis.Trained_Datasets(Inputs, Ybar_test, Ytest, NRMSE_test1, testing,
r'$NRMSE_{test} = %.3f$' % NRMSE_test1,
'Test_Dataset.png')
# Flip training and testing datasets to calculate the 2-fold crossvalidation error
Inputs, Ybar_train, Ytrain, NRMSE_train, Ybar_test, Ytest, NRMSE_test2 = TrainData(
NarmaFile, OscillatorFile, testing, training)
