ILP for Coq

Paper https://arxiv.org/abs/2411.01188

Requiremnets

• SWI-Prolog 8.4.2
• pyswip 0.2.11
• Python 3.11.4

Data

Download the dataset. Please decompose the dataset in the home directory of the project.

Explaination

• First, execute cd ilp_coq.
• To reproduce the graphs and table in Section 5, run bash shell/results.sh. The results and graphs are presented in the home directory of the project.
• To learn rules, run
  • bash shell/gen_anonym_feat_clause.sh
  • bash shell/gen_anonym_repr_clause.sh
  • bash shell/gen_origin_feat_clause.sh
  • bash shell/gen_origin_repr_clause.sh
• To evaluate in the validation dataset, run
  • bash shell/tune_anonym_feat_theory.sh
  • bash shell/tune_anonym_repr_theory.sh
  • bash shell/tune_origin_feat_theory.sh
  • bash shell/tune_origin_repr_theory.sh
• To test the learned rules in the test dataset, run
  • bash shell/train_anonym_feat_filter_theory.sh
  • bash shell/train_anonym_repr_filter_theory.sh
  • bash shell/train_origin_feat_filter_theory.sh
  • bash shell/train_origin_repr_filter_theory.sh