Skip To Content

Machine Learning Tames Monster Waves

Machine-learning and fiber laser[Enlarge image]

Combining a novel machine-learning approach with a fiber laser offers active control over the generation of rogue waves [Image: Junsong Peng]

Freak waves in the world's oceans are known for their destructive power and their disconcerting ability to appear from nowhere and then vanish without a trace. Such rogue wave events have also been found to emerge in many other complex systems, including financial models and Bose–Einstein condensates, but their unpredictability and exceptionally large amplitudes make them difficult to generate, study and control in laboratory-based systems.

Now, a research team from China, France and the UK has reportedly used a novel machine-learning approach to produce the strongest rogue waves yet recorded in an optical testbed, which in this case was based on a mode-locked fiber laser (Laser Photonics Rev., doi: 10.1002/lpor.202200470). The experimental setup can produce rogue waves with a spectral peak intensity almost 33 times larger than the average wave height, and it provides for the first time an active control mechanism for analyzing extreme wave events.

Rogue waves on demand

Fiber optic systems offer an ideal platform for probing the dynamics of unusual wave phenomena, since they allow a large number of events to be recorded within a relatively short amount of time (see “Rogue Waves of Light,” OPN, November 2015). In recent years, these optical systems have been coupled with machine-learning techniques to study stationary or repetitive wave events, such as solitons. It is much more challenging to generate rogue waves, however, because they are statistically rare and their appearance is highly unpredictable.

By using this algorithm to optimize the cavity parameters, rogue waves with controllable intensity could be triggered on demand.

In this new work, real-time spectral measurements were used as the input for a genetic algorithm—a machine-learning model based on the evolutionary principle of natural selection that is widely used to solve search and optimization problems in nonlinear optics and other complex systems. By using this algorithm to optimize the cavity parameters, rogue waves with controllable intensity could be triggered on demand.

Control in many systems

This intelligent control mechanism allowed the researchers to produce both ordinary rogue waves and super rogue waves―so named not for their strength but for their ability to appear from nowhere and disappear, just like their oceanic counterparts.

The experiments also showed that the emergence of extreme waves is accompanied by significant frequency shifts in the optical spectrum; a circulating asymmetric pulse in the cavity results in one edge of the spectrum growing into a rogue wave because of the Kerr effect, followed by the formation of an intrapulse shockwave in the time domain that ultimately leads to a pulse collapse. This suggests a new physical scenario for their formation and decay that the team confirmed with numerical simulations of the laser dynamics based on the nonlinear Schrödinger equation.

"It is reasonable to expect that the machine-learning method provided in this work can be widely implemented to control rogue waves in many different systems," said corresponding author Junsong Peng, East China Normal University, China. "Machine learning could also be used to actively control and investigate instabilities in a wide range of complex systems, avoiding the need for manual tuning to explore the richness of the underlying physics."

Publish Date: 29 November 2023

Add a Comment