OTKGE

This is the code of paper OTKGE: Multi-modal Knowledge Graph Embeddings via Optimal Transport.

Dependencies

• Python 3.6+
• PyTorch 1.0~1.7
• NumPy 1.17.2+
• tqdm 4.41.1+

Results

The results of OTKGE on WN9IMG and FBIMG are as follows.

Reproduce the Results

1. Preprocess the Datasets

First we should preprocess the datasets.

cd code
python3 process_datasets.py

Now, the processed datasets are in the data directory.

2. Reproduce the Results

CUDA_VISIBLE_DEVICES=0 python3 learn.py --dataset WN9IMG --model OTKGE_wn --rank 500 --optimizer Adagrad \
--learning_rate 1e-1 --batch_size 2000 --regularizer N3 --reg 5e-3 --max_epochs 200 \
--valid 5 -train -id 0 -save -weight

CUDA_VISIBLE_DEVICES=1 python3 learn.py --dataset FBIMG --model OTKGE_fb --rank 500 --optimizer Adagrad \
--learning_rate 1e-1 --batch_size 5000 --regularizer N3 --reg 1e-3 --max_epochs 150 \
--valid 5 -train -id 0 -save -weight