|
Listen to this story
|
Google AI has introduced an ML-generated sensory map to perceive odor. Researchers have mapped the scent of molecules to digitise the sense of smell, which acts as a potential tool to tackle global health issues.
As mentioned in their blog, the team developed a graph neural network (GNN) model in 2019, exploring thousands of examples of distinct molecules paired with the smell labels such as ‘floral’, or ‘minty’—to examine the relationship between a molecule’s structure and the probability that constituted of each smell label.
The newly introduced model called ‘Principal Odor Map’ (POM) identifies the vector representation of each odorous molecule, with the properties of a sensory map.
The first is pairs of perceptually similar odors corresponding to two nearby points in the POM. Secondly, the model enables users to predict and discover new odors and the molecules that produce them.
Google AI demonstrates that the map can be used to predict the odor properties of molecules, understand in terms of fundamental biology, and tackle global health problems.
Read the full blog here.
