The immense number-crunching capabilities artificial intelligence systems mean we can better predict the future of chaotic systems based on fewer and fewer patterns of the past – and a new algorithm is adding even more accuracy to the process.
Developed using next-gen reservoir computingThese techniques are faster and more dynamic than traditional ones. Machine learningThe new algorithm makes predictions more accurate for complex physical processes like the global weather forecast.
Calculations of these processes – known as spatiotemporal chaotic systems – can now be done in a fraction of the time, with greater accuracy, using fewer computational resources, and based on less training data.
“This is very exciting as it represents a significant advance in data processing efficiency and prediction accuracy in machine learning.” says physicist Wendson de sa BarbosaFrom Ohio State University.
Machine learning is precisely that: Computer algorithms use a discovery process for making predictions (such future weather patterns) based upon large data archives (such past weather patterns).
The reservoir computing approachThis is a way to better mimic the human brain. It feeds information into a’reservoir of randomly connected artificial neuronal neurons in an attempt to find useful patterns. These results can then be used to inform future learning cycles.
These systems have become more efficient over time. Machine learning has enabled distinct components of the predictive model can occur simultaneously. This architecture, which uses the most recent reservoir computing technology, allows algorithms to spot potential syneties in a chaotic mass of information.
They tested their approach with an atmospheric weather forecast model. The researchers were able to predict weather conditions using a laptop with Windows software. This is a significant improvement on the previous supercomputers. The calculations were performed 240,000 times faster with this particular algorithm than traditional ones.
“If one understands the equations that describe how these unique processes will evolve, then it’s possible to predict and reproduce their behavior.” says de sa Barbosa.
Machine learning algorithms can be used for predicting all types of future events and finding applications in many fields. Nothing as boring as miningNew resources for those as alarming social engineering.
As these situations become more complex, there is more to take into account, which limits computational resources. Machine learning systems can spot patterns in historical data that would have been difficult for a human eye and monitor for them repeating. They can also improve their accuracy by learning from themselves.
According to the researchers, further down the line these new and improved algorithms could be used in a wide variety of situations – such as monitoring the patterns of a heartbeat, spotting health issues that would otherwise get missed.
“Modern machine learning algorithms can be used to predict dynamical systems through learning their underlying physical rules from historical data,” says de sa Barbosa.
“Once you have sufficient data and computing power, you can make predictions using machine learning models about any real world complex system.”
The research was published in Chaos: An Interdisciplinary Journal of Nonlinear Science.
