convert lib-svm feature files into hdf5 files (tables). using static method is for consistence with multi-processing version, wh

count positive/negative samples :param hdf_data_dir: :param file_prefix: see this param in feature_to_hdf() :param n

should be accessed only in private :param X: :param y: :param batch_size: :param shuffle: :return

iterate among hdf files(blocks). when the whole data set is finished, the iterator restarts from the beginning, thus the data st

Get the min_length max_length and max_feature of data. :param file_name: The file name of input data. :return: the tuple (min

Transfer the raw data to feature data. using static method is for consistence with multi-processing version, which can not be pa

:param gen_type: 'train', 'valid', or 'test'. the valid set is partitioned from train set dynamically :param batch_size: :p

should be access only in private :param X: :param y: :return:

collect meta information, and produce hdf files if not exists :param initialized: write feature and hdf files if True

collect meta information, and produce hdf files if not exists :param initialized: write feature and hdf files if True

this method should be override :return:

summarize the data set. :return: