Quick summary:
- What I did:
- Wrote a Master’s Thesis titled “Common sense reasoning against algorithmic misinformation using Knowledge Graph Embeddings”
- Led a rotating team of 6 people through multiple projects that culminated in this research design and experimental setup
- Engineered a Python package weboftruth to train 11 KGE algorithms on any dataset with a single line of code
- Things I learnt:
- How to use intuitive representation learning for better algorithmic performance
- How to build a production-ready deep representation learning pipeline using Python and Torch
- How to write software for deep learning hardware (cuda, GPUs) etc.
- When was this?
- The latter 15 months of my time at UChicago, so May 2020 through September 2021
All code is available at the weboftruth repo. Just python -m pip install git+https://github.com/weboftruth/ to use it.
I also have Jupyter notebooks (accessible via Google Colab) that walk you through the codebase so you can run it yourself.