neupsl-ijcai23
Science Score: 18.0%
This score indicates how likely this project is to be science-related based on various indicators:
-
✓CITATION.cff file
Found CITATION.cff file -
○codemeta.json file
-
○.zenodo.json file
-
○DOI references
-
○Academic publication links
-
○Academic email domains
-
○Institutional organization owner
-
○JOSS paper metadata
-
○Scientific vocabulary similarity
Low similarity (9.1%) to scientific vocabulary
Repository
Basic Info
- Host: GitHub
- Owner: linqs
- License: apache-2.0
- Language: Python
- Default Branch: main
- Size: 189 KB
Statistics
- Stars: 0
- Watchers: 7
- Forks: 0
- Open Issues: 0
- Releases: 0
Metadata Files
README.md
Experiments for "NeuPSL: Neural Probabilistic Soft Logic" presented at IJCAI 2023
This repository covers the experiments for the paper NeuPSL: Neural Probabilistic Soft Logic presented at IJCAI 2023.
@article{pryor2023ijcai,
title = {NeuPSL: Neural Probabilistic Soft Logic},
author = {Connor Pryor* and
Charles Dickens* and
Eriq Augustine and
Alon Albalak and
William Yang Wang and,
Lise Getoor},
booktitle = {International Joint Conference on Artificial Intelligence (IJCAI)},
year = {2023}
}
Requirements
These experiments expect that you are running on a POSIX (Linux/Mac) system. The specific application dependencies are as follows: - Bash >= 4.0 - Java >= 7 - Python >= 3.7
Additionally, specific Python3 dependencies to run the exact splits are provided in requirements.txt.
If a different version of tensorflow is desired, please regenerate the data.
To install all Python3 dependencies run:
pip3 install -r ./requirements.txt
NeuPSL Experiments
To reproduce a NeuPSL experiment from the IJCAI 2023 paper simply run the following script:
./scripts/run.sh <experiment>
where <experiment> may be one of:
- citation: Citation network node classification
- mnist-addition: MNIST-Add1 and MNIST-Add2 with overlap
- vspc: Visual sudoku puzzle classification
The ./scripts/run.sh script will run NeuPSL on the specified experiment.
More specifically, it will run NeuPSL on every data setting used in the paper.
To do this it will download the data if it does not exist.
Moreover, ./scripts/run.sh will call the ./<experiment>/cli/run.sh file that fetchs the PSL .jar file from Maven central and uses it to run NeuPSL.
For individual experiments or to generate new data, please see the README in the corresponding experiment directory.
Baseline Experiments
Baseline experiments are also provided in this repository.
To reproduce baseline results for each experiment run the corresponding run script in the ./<experiment>/other-methods/<baseline>/scripts directory.
Owner
- Name: linqs
- Login: linqs
- Kind: organization
- Repositories: 36
- Profile: https://github.com/linqs
Citation (citation/README.md)
# Citation Network Node Classification ### Data Generation To generate new train/test/valid splits run the following script: ``` python3 ./scripts/create-data.py ``` After generation, specifically for the citation network node classification experiment, the neural model components for NeuPSL are pretrained with node labels before begin trained with NeuPSL. To run pretraining for the citation network node classification experiment run the following command: ``` python3 ./scripts/setup-networks.py ```
GitHub Events
Total
- Watch event: 3
- Fork event: 2
Last Year
- Watch event: 3
- Fork event: 2