Predicting seemingly random rogue waves with artificial intelligence

3 Dec 2018

Huge waves can form unexpectedly even on a tranquil sea. They appear seemingly out of nowhere and disappear just as fast. It is a fearsome phenomenon that has puzzled seamen since ancient times. Predicting these seemingly random rogue waves has so far been impossible, which has caused distress particularly for shipping.

Artificial intelligence researchers funded by the Academy of Finland have now taken a significant step towards the analysis and predictability of this phenomenon by studying optical rogue waves. The experiments were made with an optical fibre system, as rapid laser pulses operate like ocean rogue waves in specific conditions.

“Remarkably, the algorithm was shown to be capable of predicting the peak intensity of a rogue wave associated with any particular spectral measurement, even though the experiments never actually measured the rogue wave intensity directly,” says Professor Goëry Genty from Tampere University of Technology. Genty led the part of the study that was funded by Academy of Finland.

A neural network was trained to accurately pick out features that could predict the emergence of a rogue wave, even though these features were essentially invisible to the researchers.

The study was conducted by researchers from Tampere University of Technology (TUT), Finland, and Institut FEMTO-ST, Université Bourgogne-Franche Comté, France. The research at TUT was funded by the Academy of Finland.

Together, the research teams developed a technique to predict rogue waves. By using machine learning and a trained neural network, the researchers found a way to predict the intensity of rogue waves of light. The discovery enables researchers to measure the probability distribution that identifies the emergence of rogue waves.

“As well as suggesting that similar techniques can be used to analyse real-time measurements on oceanographic wave data, the results open up new perspectives in all fields of research where direct time-domain observations are difficult but where spectral data are available,” says Professor John M. Dudley from Université Bourgogne-Franche Comté.


Logo of the AIPSE ProgramThe Academy Programme for Novel Applications of Artificial Intelligence in Physical Sciences and Engineering Research (AIPSE) aims to deepen and broaden AI research expertise. AIPSE projects have been granted a total of 7 million euros in funding. The programme runs until 2021.


Last modified 18 Dec 2018
Connect with us
Facebook  Twitter YoutubeLinkedInSlideshare
SWITCHBOARD +358 295 335 000
REGISTRY +358 295 335 049
FAX +358 295 335 299
OFFICE HOURS Mon–Fri 8.00–16.15