ann,
inputs,
16,
figname=
'bpsk\\importance_heat_map_between_feature_fault_of_LSTM')
simple_net(important_vars,
important_features,
'bpsk\\lstm_simple_net.gv',
var_list=var_list,
mode_list=mode_list)
# BPSK
if 'mt' in system:
data_path = parentdir + '\\tank_systems\\data\\train\\'
mana = mt_data_manager()
mana.load_data(data_path)
mana.add_noise(20)
var_list = mana.cfg.variables
mode_list = ['normal'] + mana.cfg.faults
inputs, _ = mana.random_h_batch(batch=test_batch,
step_num=64,
prop=0.2,
sample_rate=1.0)
# CNN
if 'cnn' in network:
ann = 'mt_cnn_distill_(8, 16, 32, 64).cnn'
ann = parentdir + '\\ann_diagnoser\\mt\\train\\20db\\{}\\'.format(
args.index) + ann
important_vars = heat_map_feature_input(
ann,
inputs,
labels = torch.sum(labels * torch.Tensor([1, 2, 3, 4, 5, 6]), 1).long()
labels = labels.detach().numpy()
batch, variable, step = inputs.shape
inputs = inputs.transpose((0, 2, 1))
inputs = inputs.reshape((batch * step, variable))
labels = np.repeat(labels, step)
inputs = pca_selection.transform(inputs)
numpy2arff(inputs, labels, 'pca_test.arff', var_list, mode_list)
if args.system == 'mt':
step_len = 64
pca_selection = PCA_feature_selection(0.95)
# train
train_path = parentdir + '\\tank_systems\\data\\train\\'
mana_train = mt_data_manager()
mana_train.load_data(train_path)
mana_train.add_noise(snr)
mode_list = ['normal'] + mana_train.cfg.faults
inputs, labels = mana_train.random_h_batch(batch=batch,
step_num=64,
prop=0.2,
sample_rate=1.0)
batch, variable, step = inputs.shape
inputs = inputs.transpose((0, 2, 1))
inputs = inputs.reshape((batch * step, variable))
inputs = pca_selection.learn_from(inputs)
_, fe_num = inputs.shape
labels = np.repeat(labels, step)
var_list = ['fe' + str(i) for i in range(fe_num)]
numpy2arff(inputs, labels, 'pca_train.arff', var_list, mode_list)
# test 1