def cross_validation(self, **kwargs):
from sklearn.model_selection import KFold
kfold = KFold(n_splits=5, shuffle=True, random_state=2)
input_data, target_data = utils_conv2d.create_data_prediction(**kwargs)
count = 0
for train_index, test_index in kfold.split(input_data):
count += 1
pivot = int(0.8 * len(train_index))
input_train = input_data[train_index[0:pivot]]
input_valid = input_data[train_index[pivot:]]
input_test = input_data[test_index]
target_train = target_data[train_index[0:pivot]]
target_valid = target_data[train_index[pivot:]]
target_test = target_data[test_index]
self.input_train = input_train
self.input_valid = input_valid
self.input_test = input_test
self.target_train = target_train
self.target_valid = target_valid
self.target_test = target_test
with open("config/conv2d_gsmap.yaml") as f:
config = yaml.load(f)
config['base_dir'] = "log/conv2d/" + str(count) + '/'
self.config_model = common_util.get_config_model(**config)
self.log_dir = self.config_model['log_dir']
self.callbacks = self.config_model['callbacks']
self.model = self.build_model_prediction()
self.train()
self.test()
print("Complete " + str(count) + " !!!!")
def __init__(self, **kwargs):
self.config_model = common_util.get_config_model(**kwargs)
# load_data
self.data = utils.load_dataset(**kwargs)
self.input_train = self.data['input_train']
self.input_valid = self.data['input_valid']
self.input_test = self.data['input_test']
self.target_train = self.data['target_train']
self.target_valid = self.data['target_valid']
self.target_test = self.data['target_test']
# other configs
self.log_dir = self.config_model['log_dir']
self.optimizer = self.config_model['optimizer']
self.loss = self.config_model['loss']
self.activation = self.config_model['activation']
self.batch_size = self.config_model['batch_size']
self.epochs = self.config_model['epochs']
self.callbacks = self.config_model['callbacks']
self.seq_len = self.config_model['seq_len']
self.horizon = self.config_model['horizon']
self.input_dim = self.config_model['input_dim']
self.output_dim = self.config_model['output_dim']
self.rnn_units = self.config_model['rnn_units']
self.dropout = self.config_model['dropout']
self.latent_space = 10
self.model = self.construct_model()
self.timestep = kwargs['model'].get('timestep')
def __init__(self, **kwargs):
self.config_model = common_util.get_config_model(**kwargs)
# load_data
self.data = utils_conv2d.load_dataset(**kwargs)
self.input_train = self.data['input_train']
self.input_valid = self.data['input_valid']
self.input_test = self.data['input_test']
self.target_train = self.data['target_train']
self.target_valid = self.data['target_valid']
self.target_test = self.data['target_test']
# other configs
self.log_dir = self.config_model['log_dir']
self.optimizer = self.config_model['optimizer']
self.loss = self.config_model['loss']
self.activation = self.config_model['activation']
self.batch_size = self.config_model['batch_size']
self.epochs = self.config_model['epochs']
self.callbacks = self.config_model['callbacks']
self.seq_len = self.config_model['seq_len']
self.horizon = self.config_model['horizon']
self.model = self.build_model_prediction()