This repository contains official implementation of paper "Enhancing Inductive Numerical Reasoning in Knowledge Graphs with Relation-Aware Relative Numeric Encoding" (paper).
This repository is based on GraIL.
All the required packages can be installed by running pip install -r requirements.txt.
Our data statistics are as follows:
For traning of RRNE on Credit or Spotify, please run below code
python train.py -d $dataset -e $exp_name --input_feature rra --order_loss l2 --use_self self_1 --use_numric --self_margin 1.0 --hop $hop
For traning of RRNE on US-cities please run below code
python train.py -d USA_sparse_f -e $exp_name --input_feature rra --order_loss l2 --use_self self_1 --use_numric --self_margin 1.0 --hop 3 --self_coef 0.25
python train.py -d USA_sparse_f -e $exp_name --input_feature rra --order_loss l2 --use_self self_1 --use_numric --self_margin 1.0 --hop 4 --batch_size 4 --lr 2e-3 --self_coef 0.25
For test of RRNE, please run below code
python train.pyd -d $dataset_ind -e $exp_name --use_numeric --hop $hop
If this repository is helpful for you, please cite this paper.
@InProceedings{10.1007/978-981-96-8173-0_14,
author="Jeong, Hongjun and Jung, Heesoo and Kim, Gayeong and Kim, Juann and Kim, Ko Keun and Park, Hogun",
title="Enhancing Inductive Numerical Reasoning in Knowledge Graphs with Relation-Aware Relative Numeric Encoding",
booktitle="Advances in Knowledge Discovery and Data Mining",
year="2025",
pages="173--186",
}