Preprocess

bash init.sh

RuleTensor-TSP

• CFamily
  python RuleTensor-TSP/GraphRule_close.py -dataset=DATASET -rule_len=LEN -hc_thr=HC -sc_thr=SC -percent=0.8 -gpu=GPU

• Wiki79k and Wiki143k
  python RuleTensor-TSP/GraphRule_open.py -dataset=DATASET -rule_len=LEN -hc_thr=HC -sc_thr=SC -percent=0.8 -gpu=GPU

  -DATASET: choose the dataset in DATA/
  -LEN: set the length of rule
  -HC: set the head coverage threshold of rule
  -SC: set the standard confidence threshold of rule
  -PER: set the integrity of the dataset
  -GPU: -1 for cpu, otherwise the gpu id

  KGE-TSP

• CFamily
  python KGE-TSP/runclose.py -train -test -data=DATASET -gpu=GPU -perfix='0.8' --model=MODEL --valid_steps=STEP

• Wiki79k and Wiki143k
  python KGE-TSP/runopen.py -train -test -data=DATASET -gpu=gpu -perfix='0.8' --model=MODEL --valid_steps=STEP

  -MODEL: the choice of KGE model, ['HAKE', 'PairRE']
  -PERFIX: set the integrity of the dataset in the format of percent_
  -STEP: do valid every STEP steps

  GPHT

1. generate subgraphs

   python GPHT/run.py -dataset=DATASET -subgraph=SUBLEN -perfix=PERFIX

   -SUBLEN: set max hops of subgraph from center to edge

2. pre-train embeddings

   python GPHT/run.py -dataset=DATASET -subgraph=SUBLEN -perfix=PERFIX -batch=BATCH -pretrain -desc=DESC

3. train the model

   python GPHT/run.py -dataset=DATASET -perfix=PERFIX -lr=LR -restore=RESTORE -batch=1 -epoch=EPOCH -valid_epochs=STEP -score_func=MODEL -minconf=MINCONF

   -LR: a little scale number for learning rate, like 0.00003 or less
   -MINCONF: selecting the final predicted triples

4. predict triples(in KGE-TSP)

   • CFamily
     python KGE-TSP/runclose.py -train -test -data=DATASET -gpu=0 -perfix='0.8' -testGNN "EXPS/CFamily/toKGE_XXX.pt" -model=MODEL

   • Wiki143k and Wiki79k
     python KGE-TSP/runopen.py -train -test -data=DATASET -gpu=0 -perfix='0.8' -testGNN "EXPS/DATASET/toKGE_XXX.pt" -model=MODEL -valid_steps=STEP

Acknowledgement

We refer to the code of HAKE、PairRE and CompGCN. Thanks for their contributions.