T

type-constrained_representation_learning

Code and datasets used in the publication: Type-Constrained Representation Learning in Knowledge Graphs. Denis Krompass, Stephan Baier and Volker Tresp Proceedings of the 13th International Semantic Web Conference (ISWC). 2015

Name Last Update
type_constrained_representation_learning Loading commit data...
.gitignore Loading commit data...
LICENSE.txt Loading commit data...
README.txt Loading commit data...
This folder contains the code used for the experimental results shown in the 
paper "Type-Constrained Representation Learning in Knowledge Graphs" (ISWC 2015)

Notes:
./experiments/experiments.py is a script to manage and start various experiments 
for hyperparameter tuning or evaluation on a hold out set on DBpedia-Music, 
Freebase-150k, YAGO-195k for RESCAL, TransE or the GoogleKnowledgeVault Neural 
Network.

You have to add the project folder ("type-constrained_representation_learning")
to your $PYTHONPATH.

execute
$ cd experiments
$ python experiments.py -h 
to display more details on the execution of the script.

All experiments and models are build around a central data structure 
(CooTripleTensor, ./data_structures/triple_tensor/coo_triple_tensor.py) that 
allows the efficient storage and indexing of larger triple stores represented
as very sparse 3D tensors (entities x entities x relation-types).

The implementations of the models can be found in ./models

python packages required:
-------------------------
numpy 
scipy
sklearn
joblib
theano

We compile numpy with OpenBlas:
https://github.com/xianyi/OpenBLAS


================================================================================
Please cite:
============
Denis Krompass, Stephan Baier and Volker Tresp
Type-Constrained Representation Learning in Knowledge Graphs
International Semantic Web Conference 2015

Bibtex:
@inproceedings{KrompassISWC2015,
  author    = {Denis Krompa\ss and Stephan Baier and Volker Tresp},
  title     = {Type-Constrained Representation Learning in Knowledge Graphs},
  booktitle = {Proceedings of the 13th International Semantic Web Conference (ISWC)},
  year      = {2015}
}
================================================================================