ALTeGraD-2023 Data Challenge
Kaggle challenge: https://www.kaggle.com/competitions/altegrad-2023-data-challenge/overview
Team: Baku incorporated
This project is dedicated to aligning textual queries with graph representations of molecules, prompting a systematic investigation into various methodologies. The core approach involves jointly training specialized encoders, one designed for processing textual data and another finely tuned to molecular structures. Utilizing contrastive learning, the model aims to map similar text-molecule pairs closely in the learned representation space while enforcing separation for dissimilar pairs
