Russian researchers are using deep learning neural networks to sniff out potential scent-based threats. The technique is a bit dense (as anything with neural nets tends to be), but the gist is that the electronic "nose" can remember new smells and recognize them after the fact.
When the sensor detects a smell, an AI takes over and checks it against a database of known scents for "the closest similar smell determined by the smallest Hamming distance to any know code," HSE writes. If it can't find a match, the sensor will identify the scent as being new.
The difference between this and other scent prediction/identification tech -- like the crowdfunded one from this February -- is that it can sense more than one scent at a time. Useful for, say, gas mixtures.
"Essentially, we want to teach the device to discriminate between hazardous and non-hazardous gas mixtures and memorize them fast," MIEM HSE professor Vladimir Kulagin says.
At the least, it could greatly benefit folks working in mines or enclosed spaces as sort of a digital canary, alerting folks when an unsafe gas has entered the area.