Please consider the following steps to get the results of Table 2.

1. download the data sets from the links that we support in the paper.

2. process the data sets to the pickle format. (We give an example in ``./datasets/pickle-format/eurlex4k.pkl'')

3. split each data set into five folds for cross-validation. (You can consider to use the script ``cv_split.sh'')

4. pre-build the trees by ``construct_trees.sh'' to avoid duplicated label tree construction.

5. execute train_tree_cv.sh and train_tree_with_full_tr_data.sh

6. execute test_tree_cv.sh to have the results of Table 2. 
  If you want to record the log, please consider the commend ``$ bash test_tree_cv.sh > results.log''


