In the popular myth, lemmings are said to run off cliffs to their doom. However, this is not true. Nevertheless, many real-world systems experience critical points and abrupt disasters, such as stock market crashes and power grid failures. Critical points can be values of certain system parameters that mark the transition to instability.
A new method has been developed to determine when a system is close to a critical point. Previous research has shown that systems tend to slow down and become more variable near critical points. However, these indicators do not work well for noisy systems. The researchers searched through thousands of methods and found a few that performed well even in noisy systems. Based on these methods, they developed a simple new recipe called RAD (Rescaled AutoDensity) for predicting critical points.
The researchers tested their method on recordings of brain activity in mice. They found that certain areas of the brain showed stronger signs of being close to a critical point than others. This suggests that the brain may use critical points to support its computational abilities. Being far from a critical point makes neural activity stable and supports efficient processing of basic visual features. However, being close to a critical point may provide a longer “memory” to support more complex computations.
Understanding systems near critical points can have important applications in finance and medicine. It can help predict sudden changes and potentially prevent disasters such as seizures or financial crashes.