Machine Learning for predicting future in chaotic systems

Machine learning is a part of the computer field that deals with using algorithms for obtaining certain data. It has shown many of its applications in many sectors such as for the recognization of images or voice.

Keywords: Machine Learning, spatiotemporal chaotic systems, algorithm, next-generation reservoir computing, atmospheric weather, accurate, training data points.

Now there is a new development in the field of machine learning. Researchers at The Ohio State University have discovered a new way of predicting the behavior of spatiotemporal chaotic systems. These systems mainly make predictions for changes in Earth's weather. Generally, it is very challenging for scientists to predict these changes. Now, the new technique of ML called next-generation reservoir computing will give an easy way to it.

Research is published in the journal Chaos. In this research, the new and efficient algorithm when combined with next-generation reservoir computing learned spatiotemporal chaotic systems in less time compared to the traditional ways.

A challenging problem that was researched in the past was predicting the behavior of an atmospheric weather model. The researchers now tried to test the same by using the new algorithm. The new algorithm is more precise and it needs 400 to 1,250 times less training data to make predictions compared to the traditional ML algorithms. 

The researchers made use of a Windows 10 laptop for making predictions using the new algorithm. It was roughly 240,000 times faster than the traditional ML algorithms. Also, the computation method is less complex in the new algorithm whereas the complexity in the traditional method is more, also it requires a supercomputer.

Wendson De Sa Barbosa, lead author and a postdoctoral researcher in physics at Ohio State said, "This is very exciting, as we believe it's a substantial advance in terms of data processing efficiency and prediction accuracy in the field of machine learning. Learning to predict these extremely chaotic systems is a 'physics grand challenge,' and understanding them could pave the way to new scientific discoveries and breakthroughs. If one knows the equations that accurately describe how these unique processes for a system will evolve, then its behavior could be reproduced and predicted."

The point of view of the researcher justified that the prediction of simple movements such as the swing position of the clock is easier as it can be predicted by determining its current position and velocity. But in the case of predicting Earth's weather, it is very complex as many variables are used for determining the chaotic behavior.

Thus for predicting accurately the entire system, the scientists require detailed information on each variable, the model equations that describe the relationships among the variables. Due to this, the traditional algorithm used 50,000 historical training data points as examined in past work. But using the new algorithm it can be reduced to only 400, still achieving accurate results.