f08653c95fbbff7db73244b97839caf10883a7ff,openrec/modules/interactions/ns_eu_dist.py,NSEuDist,_build_training_graph,#NSEuDist#,39

Before Change

                                        reduction_indices=1,
                                        keep_dims=True, name="l2_user_pos")
            l2_user_neg = tf.reshape(tf.reduce_sum(tf.square(tf.subtract(tmp_user, self._n_item)),
                                                   reduction_indices=[1, 2], name="l2_user_neg"), [-1,1])
            pos_score = l2_user_pos + self._p_item_bias
            neg_score = l2_user_neg + tf.reduce_sum(self._n_item_bias, reduction_indices=[1, 2])
            self._loss = tf.reduce_sum(self._weights * tf.maximum(self._margin*self._neg_num + pos_score*self._neg_num - neg_score, 0))

After Change

                                        reduction_indices=2, 
                                        name="l2_user_neg")
            pos_score = l2_user_pos + tf.tile(self._p_item_bias, [1, self._neg_num])         // shape=(2000, self._neg_num)
            neg_score = l2_user_neg + tf.reduce_sum(self._n_item_bias, reduction_indices=2)  // shape=(2000, self._neg_num)
            self._loss = tf.reduce_sum(self._weights * tf.maximum(self._margin + pos_score - neg_score, 0))

    def _build_serving_graph(self):

Instances