Resumen de Computers predict how chemicals smell
Chemists have never been able to predict how molecules will smell, except in a few special cases. Andreas Keller and Leslie Vosshall at Rockefeller University in New York City wondered if crowdsourced machine learning could help. They asked 49 volunteers to rate the odor of hundreds of chemicals, according to how intense or pleasant each was, and how well it matched 19 other descriptors. Then they released the ratings for 407 of these chemicals, along with 4,884 chemical structure details for each, and asked people to use this to develop machine-learning algorithms. In all, 22 teams entered the challenge. They were evaluated using 69 chemicals whose volunteer ratings hadn't been released. The best algorithms proved far more accurate than previous efforts at predicting the volunteers' descriptions of the test chemicals.
