:param query_batch: 查询输入 :param doc_batch: 文档输入 :param label_batch: 查询和文档对应的相关性label :return:

计算一次迭代过程 :param sess: :param query_input: :param doc_input: :param labels: :return:

Split a word with lenth of self.n_gram_size :param self: :param word: word to be splited by n_gram len :return:

模型入口 :param sess: :param query_input: :param doc_input: :param labels: :return:

Get n_gram terms and mapping them to indexes :return:

计算log_loss, 也就是交叉熵 :return:

用极大似然的方法计算, 正例的条件概率 计算相关文档的loss, gama经验值也用来学习 :return:

建立整个模型，分成两端的网络，query端和doc端的

采用梯度下降方式学习 :return:

Get n_gram counting term matrix :param sentences: sentences to be handled to get n_gram term counting matrix :param is_du

测试函数 :return:

测试函数 :return:

paras: hash_tokens_nums: word hash后词的个数 dnn_layer_nums: dnn的层数 dnn_hidden_node_nums: dnn的

:param words: origin vacabulary :param n_gram: number of letters to make a n_gram term :param marks: the character added in

Calculate cosine distance about vector x and y :param x: np.array liked, a vector, one dimension :param y: np.array liked, a vector, one

:param orig_X: :param orig_y: :param orig_label: :param batch_size: :param shuffle: :return:

计算预测过后的查询与文档的相关性 :param sess: :param query: :param doc: :param labels: :return: