DataLoad

Load and pre-process the dataset for neural network training, including data cleaning, Box-Cox transformation and normalization.

DataLoad.normInfo(input_mean, label_mean, input_std, label_std)

Store mean and standard deviation of input/label dataset into an argparser-like object.

Parameters

• input_mean (numpy.ndarray) – Input dataset mean.

• label_mean (numpy.ndarray) – Label datatset mean.

• input_std (numpy.ndarray) – Input dataset standard deviation.

• label_std (numpy.ndarray) – Label dataset standard deviation.

Returns

norm – Normalization infomation containing mean and standard deviation.

Return type

easydict.Dict

DataLoad.loadData(args)

Load dataset and excute Box-Cox transformation and normalization.

Parameters

args (argparse.Parser) – Hyper-parameters.

Returns

• input_train (torch.Tensor) – Input dataset for training.

• label_train (torch.Tensor) – Label dataset for training.

• input_valid (torch.Tensor) – Input dataset for validation.

• label_valid (torch.Tensor) – Label dataset for validation.

• norm (argparse.Parser or easydict.EasyDict) – Normalization information.