SML Project1 Team 22

Team Members

• Yichao Xu - 1045184
• Xinglin Qiang - 1153086
• Shaohua Liu - 1150336

Kaggle Leaderboard

AUC Score: 0.92258

Environment Setup

From Debian 9 scratch

System dependencies

• glibc ≥ 2.27 (add deb http://ftp.debian.org/debian sid main to /etc/apt/sources.list)
• CUDA ≥ 11.0
• Python ≥ 3.6

Python dependencies (recommend to manage the packages by Anaconda)

• jupyterlab latest
• TensorFlow 2.3.0
• Pytorch 1.3.1
• GraphVite 0.2.2 Package Link

Example Invocation

Run command:

graphvite run my_config.yml

Here's the config file that we used for the final submission where you can tune the hyper-parameters:

./my_config.yml

The graph embedding file (in pickle format) will be generated into the directory specified in the config file.

Then you can use the code in eval/evaluate.ipynb to generate the formatted probability CSV file for Kaggle submission.

Scripts Roles

• my_config.yml - Config settings for running GraphVite.
• adj2edge.ipynb - Transforming adjacent list form to edge list form.
• evaluate.ipynb - Calculating AUC scores and Kaggle competition results.
• split_dataset.ipynb - Randomly splitting the raw data into 8:2 dataset, while generating fake edges in the 20% dataset.
• eval.ipynb - Making evaluation on the embeddings.
• logistic_reg.ipynb - Trying Logitic Regression based on the embeddings.