ichise-laboratory / tuckerdncaching

We extend the idea of reducing false negatives by adopting a Tucker decomposition representation to enhance the semantic soundness of latent relations among entities by introducing a relation feature space.

TuckerDNCaching

Project Setup

To set up the project, follow these steps:

1. Clone the repository:

git clone https://github.com/ichise-laboratory/tuckerdncaching.git

2. Install the required dependencies:

pip install -r requirements.txt

3. Run the project:

python examples/complex_fb15k237.py

Best Results

Comparison of state-of-the-art negative sampling methods on WN18, WN18RR, FB15K and FB15K237 datasets Note that results for the MR metric and all metric results for the DistMult KGE model for all datasets are not available for KSGAN. Metric results for all KGE models for the FB15K dataset are also not available for KSGAN as the original did not include them.

Paper Citation

If you find this research helpful, please consider citing our paper:

@article{tiroshan2022tuckerdncaching,
	title        = {Tucker{DNC}aching: high-quality Negative Sampling with Tucker Decomposition},
	author       = {Madushanka, Tiroshan and Ichise, Ryutaro"},
	journal      = {Journal of Intelligent Information Systems},
	volume       = {60},
        number       = {3},
	pages        = {},
	url          = {https://doi.org/10.1007/s10844-023-00796-y}
	doi          = {10.1007/s10844-023-00796-y}
}