Need help with kbc?
Click the “chat” button below for chat support from the developer who created it, or find similar developers for support.

About the developer

facebookresearch
164 Stars 26 Forks Other 15 Commits 0 Opened issues

Description

Tools for state of the art Knowledge Base Completion.

Services available

!
?

Need anything else?

Contributors list

# 238,316
Shell
PHP
pytorch
Deep le...
15 commits

Knowledge Base Completion (kbc)

This code reproduces results in Canonical Tensor Decomposition for Knowledge Base Completion (ICML 2018).

Installation

Create a conda environment with pytorch cython and scikit-learn :

conda create --name kbc_env python=3.7
source activate kbc_env
conda install --file requirements.txt -c pytorch

Then install the kbc package to this environment

python setup.py install

Datasets

To download the datasets, go to the kbc/scripts folder and run:

chmod +x download_data.sh
./download_data.sh

Once the datasets are download, add them to the package data folder by running :

python kbc/process_datasets.py

This will create the files required to compute the filtered metrics.

Running the code

Reproduce the results below with the following command :

python kbc/learn.py --dataset FB15K --model ComplEx --rank 500 --optimizer
Adagrad --learning_rate 1e-1 --batch_size 1000 --regularizer N3 --reg 1e-2
 --max_epochs 100 --valid 5

Results

In addition to the results in the paper, here are the performances of ComplEx regularized with the weighted N3 on several datasets, for several dimensions. We use an init scale of 1e-3, a learning rate of 0.1, a batch size of 1000 and 100 max epochs unless specified otherwise. We use the Adagrad optimizer.

FB15k

For rank 2000 : learning rate 1e-2, batch-size 100, max epochs 200.

| rank | 5|25|50|100|500|2000| |------------|--|--|--|---|---|----| | MRR | 0.36|0.61|0.78|0.83|0.84|0.86 | | [email protected] | 0.27|0.52|0.73|0.79|0.80|0.83 | | [email protected] | 0.41|0.67|0.81|0.85|0.87|0.87 | | [email protected] | 0.55|0.77|0.86|0.89|0.91|0.91 | | | | | | | | | | | | | | | | | | reg | 1e-5|1e-5|1e-5|7.5e-4|1e-2|2.5e-3 | | #Params | 163k|815k|1.630M|3.259M|1.630M|65.184M |

WN18

Max Epochs : 20

| rank | 5|8|16|25|50|100|500|2000 | |------------| -|-|-|-|-|-|-|- | | MRR | 0.19|0.45|0.92|0.94|0.95|0.95|0.95|0.95 | | [email protected] | 0.14|0.37|0.91|0.94|0.94|0.94|0.94|0.94 | | [email protected] | 0.20|0.50|0.93|0.94|0.95|0.95|0.95|0.95 | | [email protected] | 0.29|0.60|0.94|0.95|0.95|0.95|0.96|0.96 | | | | | | | | | | | | | | | | | | | | | | reg | 1e-3|5e-4|5e-4|1e-3|5e-3|5e-2|5e-2|5e-2 | | #Params | 410k|656k|1.311M|2.049M|4.098M|8.196M|40.979M|163.916M|

FB15K-237

Batch Size : 100 (1000 for rank 1000)

| rank | 5|25|50|100|500|1000|2000 | |------------| -|-|-|-|-|-|- | | MRR | 0.28|0.33|0.34|0.35|0.36|0.37|0.37 | | [email protected] | 0.20|0.24|0.25|0.26|0.27|0.27|0.27 | | [email protected] | 0.31|0.36|0.37|0.39|0.40|0.40|0.40 | | [email protected] | 0.44|0.51|0.52|0.54|0.56|0.56|0.56 | | | | | | | | | | | | | | | | | | | | reg | 5e-4|5e-2|5e-2|5e-2|5e-2|5e-2|5e-2 | | #Params | 150k|751k|1.502M|3.003M|15.015M|30.030M|60.060M |

WN18RR

Batch Size : 100 (1000 for rank 8)

| rank | 5|8|16|25|50|100|500|2000 | |------------| -|-|-|-|-|-|-|- | | MRR | 0.26|0.36|0.42|0.44|0.46|0.47|0.49|0.49 | | [email protected] | 0.20|0.38|0.39|0.41|0.43|0.43|0.44|0.44 | | [email protected] | 0.29|0.38|0.42|0.45|0.47|0.49|0.50|0.50 | | [email protected] | 0.36|0.41|0.46|0.49|0.52|0.56|0.58|0.58 | | | | | | | | | | | | | | | | | | | | | | reg | 5e-4|5e-4|5e-2|1e-1|1e-1|1e-1|1e-1|1e-1 | | #Params | 410k|655k|1.311M|2.048M|4.097M|8.193M|40.975M|163.860M |

YAGO3-10

| rank | 5|16|25|50|100|500|1000 | |------------| -|-|-|-|-|-|- | | MRR | 0.15|0.34|0.46|0.54|0.56|0.57|0.58 | | [email protected] | 0.10|0.26|0.38|0.47|0.49|0.50|0.50 | | [email protected] | 0.16|0.37|0.50|0.58|0.60|0.62|0.62 | | [email protected] | 0.25|0.50|0.60|0.67|0.69|0.71|0.71 | | | | | | | | | | | | | | | | | | | | | reg | 1e-3|1e-4|5e-3|5e-3|5e-3|5e-3|5e-3 | | #Params | 1.233M|3.944M|6.163M|12.326M|24.652M|123.262M|246.524M|

License

kbc is CC-BY-NC licensed, as found in the LICENSE file.

We use cookies. If you continue to browse the site, you agree to the use of cookies. For more information on our use of cookies please see our Privacy Policy.