Scientists have taught computers to better find distant nuclear explosions

According to a new method developed by a scientist at the University of Alaska Fairbanks, computers canbe

trained to better detect distant nuclear and chemical explosions and volcanic eruptions by learning based on signals fromman-made explosionsDiscuss

The work, led by UAF Geophysical Institute postdoctoral fellow Alex Witsil, was published in Geophysical Research Letters.

Witsil his colleagues created a librarysynthetic signals of infrasonic explosions to train computers to recognize the source of the infrasonic signal. Infrasound is too low a frequency to be heard by humans and travels farther than high frequency audible waves.

"We used simulation software to create 28,000 synthetic infrasound signals that, although computer-generated, could hypotheticallybe recorded by infrasound microphones installed hundreds of kilometers away from the explosion," Witsil said.