Learning to smell and helping machines do the same

Ambuj Tewari, Statistics | 2022-2023


STATS 315 is a new introductory course on deep learning that teaches students the basics of a powerful methodology that has tremendous societal impact. However, all the datasets we use in the course currently are already curated and tend to be focused on images or text. This project will help students learn to smell odorants and create a novel dataset with the help of their own participation. Then they will use deep learning techniques to predict odorant perceptual qualities using molecular structure. This project encourages active learning, enhances student agency, and helps students appreciate a sense which is often neglected in many cultures. Often people only realize the value of the human ability to smell only when it is lost (as we saw during the COVID-19 pandemic).