Reproducability-Study

Interpretable Clustering on Dynamic Graphs with Recurrent Graph Neural Networks

This repository is the official implementation of Interpretable Clustering on Dynamic Graphs with Recurrent Graph Neural Networks(https://arxiv.org/abs/2012.08740).

Requirements

To install requirements:

setup run python pip install tensorflow-gpu pip install tensoflow_addons pip install numpy pip install spektral

To run non model main:

setup run python 3.6 pip install dgl-cu101 pip install dynamicgem pip install keras==2.2.4 pip install torch pip install --user scipy==1.4.1 pip install sklearn

These are seperate because dynamicgem has a lot of specific dependencies that make it incompatible with the original environment

Training

To train the model(s) in the paper, run:

train main.ipynb

To train models which don't need to be trained in the paper, run: train non model main.ipynb

Evaluation

To evaluate the model(s) in the paper, run:

train main.ipynb

To evaluate models which don't need to be trained in the paper, run: train non model main.ipynb

Pre-trained Models:

Pre-trained Models are not available yet. Models take around 5 minutes to train.

Results

Our model achieves the following performance on :

DBPL3(https://paperswithcode.com/paper/interpretable-clustering-on-dynamic-graphs)

| Model name | Accuracy | AUC | F1 | | ------------------ |---------------- | --------------- | --------------- | | Graphsage | 77.56% | 86.46% | 69.77% | | GCN | 78.34% | 89.12% | 69.45% | | GAT | 78.17% | 87.76% | 68.81% | | Dynaernn | 45.69% | 51.57% | 52.34% | | Spectral | 76.22% | 50.26% | 66.63% | | GCNLSTM | 77.48% | 86.5% | 70.56% | | RNNGCN | 77.83% | 88.28% | 69.26% | | TRNNGCN | 77.84% | 87.39% | 69.51% |

DBPL5(https://paperswithcode.com/paper/interpretable-clustering-on-dynamic-graphs)

| Model name | Accuracy | AUC | F1 | | ------------------ |---------------- | --------------- | --------------- | | Graphsage | 66.5% | 80.49% | 58.9% | | GCN | 68.5% | 87.67% | 56.31% | | GAT | 68.74% | 86.97% | 56.67% | | Dynaernn | 37.36% | 51.06% | 41.63% | | Spectral | 67.3% | 54.16% | 50% | | GCNLSTM | 67.68% | 84.57% | 57.66% | | RNNGCN | 68.55% | 85.99% | 57.85% | | TRNNGCN | 68.65% | 85.85% | 57.58% |

Reddit(https://paperswithcode.com/paper/interpretable-clustering-on-dynamic-graphs)

| Model name | Accuracy | AUC | F1 | | ------------------ |---------------- | --------------- | --------------- | | Graphsage | 28.8% | 56.37% | 16.38% | | GCN | 29.24% | 55.93% | 18.75% | | GAT | 31.85% | 55.92% | 15.54% | | Dynaernn | 29.16% | 52.44% | 28.98% | | Spectral | 32.02% | 50.14% | 15.93% | | GCNLSTM | 31.23% | 56.7% | 20.93% | | RNNGCN | 31.85% | 55.92% | 15.54% | | TRNNGCN | 30.96% | 56.18% | 17.57% |

Brain(https://paperswithcode.com/paper/interpretable-clustering-on-dynamic-graphs)

| Model name | Accuracy | AUC | F1 | | ------------------ |---------------- | --------------- | --------------- | | Graphsage | 64.93% | 91.29% | 91.29% | | GCN | 21.12% | 67.62% | 12.56% | | GAT | 39.81% | 82.6% | 33.18% | | Dynaernn | 26.28% | 58.61% | 26.01% | | Spectral | 36.36% | 64.18% | 36.68% | | GCNLSTM | 41.52% | 85.1% | 40.1% | | RNNGCN | 30.04% | 76% | 24.66% | | TRNNGCN | 21.94% | 66.42% | 15.58% |

📋 Include a table of results from your paper, and link back to the leaderboard for clarity and context. If your main result is a figure, include that figure and link to the command or notebook to reproduce it.

Contributing

📋 Pick a licence and describe how to contribute to your code repository.