Skip to content
Hyperbolic Disk Embeddings for Directed Acyclic Graphs (ICML 2019)
Python
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
data/maxn
models
.gitignore
.python-version
LICENSE
README.md
__init__.py
config.py
evaluations.py
named_tasks.py
relations.py
requirements.txt
run.py
split_wordnet_data.py
target.py
tasks.py
utils.py

README.md

Disk Embeddings

Anonymized version of implementation of Disk Embeddings. Full version will be available online in camera-ready version.

This code is derived from Ganea et al. 2018, and there might be comments by them or their collaborators, however, the authors are not aware of it.

We conducted the experiments with python 3.6.6, Ubuntu 18.04.1 LTS in AWS EC2 c5.18xlarge instance.

Baseline methods

Evaluated baseline methods are as follows:

  • Poincare Embeddings (Nickel et.al., 2017)
  • Order Embeddings (Vendrov et.al., 2016)
  • Hyperbolic Entailment Cones (Ganea et.al., 2018)

Data Preparation

  • We used WordNet data already processed by Ganea et al. 2018.
  • Preprocess on Hep-Th was made in notebook/Citation.ipynb

Evaluation

For single worker,

python run.py RunAll --local-scheduler

For parallelization, luigid scheduler is required to be running.

python run.py RunAll --workers=32
You can’t perform that action at this time.