RMNA: A Neighbor Aggregation-Based Knowledge Graph Representation Learning Model Using Rule Mining

RMNA: A Neighbor Aggregation-Based Knowledge Graph Representation Learning Model Using Rule Mining

Our code is based on Learning Attention-based Embeddings for Relation Prediction in Knowledge Graphs

This README is also based on it.

This repository contains a Pytorch implementation of RMNA. We use AMIE to obtains horn rules. RMNA is a hierarchical neighbor aggregation model, which transforms valuable multi-hop neighbors into one-hop neighbors that are semantically similar to the corresponding multi-hop neighbors, so that the completeness of multi-hop neighbors can be ensured.

Requirements

• conda
• Java

Please download miniconda from above link and create an environment using the following command:

    conda env create -f pytorch35.yml

Activate the environment before executing the program as follows:

    source activate pytorch35

Dataset

We used two datasets for evaluating our model. All the datasets and their folder names are given below.

• Freebase: FB15k-237
• Wordnet: WN18RR

Rule Mining and Filtering

In the AMINE+ folder, we can generate mining rules by using the following command:

    java -jar amie_plus.jar [TSV file]

Without additional arguments AMIE+ thresholds using PCA confidence 0.1 and head coverage 0.01. You can change these default settings. See AMIE. The available files generated and processed by AMIE are placed in the folder of the corresponding dataset named new_triple.

Training

Parameters:

--data: Specify the folder name of the dataset.

--epochs_gat: Number of epochs for gat training.

--epochs_conv: Number of epochs for convolution training.

--lr: Initial learning rate.

--weight_decay_gat: L2 reglarization for gat.

--weight_decay_conv: L2 reglarization for conv.

--get_2hop: Get a pickle object of 2 hop neighbors.

--use_2hop: Use 2 hop neighbors for training.

--partial_2hop: Use only 1 2-hop neighbor per node for training.

--output_folder: Path of output folder for saving models.

--batch_size_gat: Batch size for gat model.

--valid_invalid_ratio_gat: Ratio of valid to invalid triples for GAT training.

--drop_gat: Dropout probability for attention layer.

--alpha: LeakyRelu alphas for attention layer.

--nhead_GAT: Number of heads for multihead attention.

--margin: Margin used in hinge loss.

--batch_size_conv: Batch size for convolution model.

--alpha_conv: LeakyRelu alphas for conv layer.

--valid_invalid_ratio_conv: Ratio of valid to invalid triples for conv training.

--out_channels: Number of output channels in conv layer.

--drop_conv: Dropout probability for conv layer.

How to run

When running for first time, run preparation script with:

    $ sh prepare.sh

• Freebase

    $ python3 main.py --data ./data/FB15k-237/ --epochs_gat 2000 --epochs_conv 150  --get_2hop True --partial_2hop True --batch_size_gat 272115 --margin 1 --out_channels 50 --drop_conv 0.3 --output_folder ./checkpoints/fb/out/
  

• Wordnet

    $ python3 main.py --data ./data/WN18RR/--epochs_gat 3600 --epochs_conv 150 --get_2hop True --partial_2hop True