05.3.18

AI Prediction System Extends Forecasts of Chaotic Systems

Chaotic systems are hard to predict because—well—they’re chaotic.

However, given that humanity is constantly at the mercy of the vagaries of chaotic systems—such as weather, solar storms and sometimes even road traffic—scientists are striving to develop methods to forecast such disorganized events. The latest breakthrough on this front comes from researchers leveraging AI prediction technology to forecast the evolution of chaotic systems for unprecedented lengths of time, according to Quanta Magazine.

Chaos theorist Edward Ott and a team at the University of Maryland have devised a machine-learning algorithm that can learn the dynamics of chaotic system called the Kuramoto-Sivashinsky equation. Often used as a test bed for studying certain types of chaotic systems, Kuramoto-Sivashinsky acts like a wave of flame travelling through combustible matter.

By training itself on the past behavior of the equation, the algorithm was able to predict the progress of the flame eight times further into the future than previous methods could.
“This is really very good,” Holger Kantz, a chaos theorist at the Max Planck Institute for the Physics of Complex Systems in Dresden, Germany, told Quanta. “The machine-learning technique is almost as good as knowing the truth, so to say.”

Significantly, the algorithm is not using the Kuramoto-Sivashinsky equation itself to conduct AI prediction. Instead, it uses existing data previously generated by the equation to produce a forecast of what it will do next.

This is critical, given that there are no known equations to describe many chaotic systems. The lack of underlying equations makes it hard to model and predict such systems, Quanta noted.

The AI prediction algorithm eliminates the need to develop complex equations, instead just ingesting data to extrapolate future trends.

“This paper suggests that one day we might be able perhaps to predict weather by machine-learning algorithms and not by sophisticated models of the atmosphere,” Kantz said.
The algorithm also could predict other chaotic events, such as predicting rogue waves and earthquakes.

“I think it’s not only working in the example they present but is universal in some sense and can be applied to many processes and systems,” said chaos expert Ulrich Parlitz of the Max Planck Institute for Dynamics and Self-Organization.

In addition to weather forecasting and other natural events, the algorithm has many other potential uses, including monitoring patients for signs of heart attacks and could monitor the brain activity of patients for indications of nerve impulses.