The technique combines conventional AI with a statistical approach known as Bayesian reasoning.

According to a new study, “Black Swan” disasters including earthquakes, pandemics, and hurricanes may be predicted more easily by using conventional artificial intelligence along with a statistical method known as Bayesian reasoning. This approach does not require extensive historical data to train AI algorithms particularly when such events are very rare. (Spectrum.ieee.org)

In Bayesian reasoning, prior knowledge helps determine the chances that an uncertain choice might be correct.

Since the data relating to Black Swan events is very scarce, this combination of AI and Bayesian reasoning identifies the most useful data points for making reliable forecasts of such disastrous events.

“Typically, AI needs plentiful data to be successful,” says study lead author Ethan Pickering. “With our approach, only a small amount of carefully selected data is needed to provide accurate and reliable results.” Pickering is a machine learning research scientist now at Bayer Crop Sciences in Cambridge, Massachusetts.

This method also bypassed the previous, cumbersome, equation-based method by simply using data and what is known as a deep neural network.  The latter mimics the neurons in the human brain.

“AI provides a flexible opportunity to skip over (this) fundamental equation-based challenge and base our understanding of these systems on observed truth,” Pickering says.  However, “using an ensemble of deep neural operators results in uncertainty quantification that allows our approach to work well,” he adds.

The new strategy successfully estimated the chances of rare black swan occurrences such as a spike in the number of cases during a pandemic, the risk of a ship cracking in half due to stress, and a normal wave that could turn into a dangerous rogue wave.

“We were simply surprised at how well our approach worked for every system we tried, including many not described in the paper, as well as very complex systems,” Pickering says.

He adds that the system could be used to predict “the likely and catastrophic scenarios that society must prepare for in 2023” due to El Niño.

Related Story:  AI Helps Researchers Understand Extreme Weather In The Midwest