Can AI Sniff the Roses, Predict the Smell?

Spent some time in examining artificial olfaction for the coffee business. Here we are looking at predicting what something smells or tastes like.   The general approach of predictive machine learning is well described below.  The broad idea of 'digital scent' has been broadly mentioned in this blog.

Using AI to Smell the Roses
UC Riverside News
Iqbal Pittalwala

Researchers at the University of California, Riverside (UCR) used machine learning to understand what a chemical smells like to humans. UCR's Anandasankar Ray and Joel Kowalewski developed a technique for a computer to learn chemical features that activate known human odorant receptors (ORs), then screened about 500,000 compounds for binding ligand molecules for 34 ORs. The team then concentrated on whether their OR activity-estimating algorithm could predict diverse perceptual qualities of odorants, and learned that such activity accurately predicted 146 distinct chemical percepts. Only a few ORs were required for prediction. Said Ray, "The machine learning algorithm can eventually predict how a new chemical will smell, even though we may initially not know if it smells like a lemon or a rose."  ... "