Physics Discovery Predicts When a System Is Nearing a Critical Point
As the myth goes, lemmings sometimes follow each other off cliffs to their doom. Imagine joining your fellow lemmings on a sunny mountain hike, enjoying the day—until suddenly, you plunge into the air, and everything goes black.
The cliff’s edge represents what scientists refer to as a “critical point,” where a system’s behavior (like a group of lemmings) abruptly shifts from one state (happily running) to a drastically different state (plummeting), often with disastrous consequences.
Although lemmings don’t actually leap off cliffs, many real-world systems do encounter critical points leading to sudden catastrophes, such as stock market crashes, power grid failures, and tipping points in climate systems and ecosystems.
Critical Points as System Parameters
Critical points aren’t always literal locations or moments. They can also be specific values of a system parameter—like investor confidence, environmental temperature, or power demand—that signal a transition to instability.
Is it possible to detect when a system is nearing a critical point and intervene to prevent a collapse? What indicators in a stock market or ecosystem could help us gauge its proximity to such a threshold?
We have developed a new method to predict these critical points in real-world systems, and our findings are published this week in Physical Review X.
Systems “Slow Down” and Become Unpredictable Near Critical Points
Earlier research has demonstrated that systems often “slow down” and become more unpredictable as they approach critical points. In the context of a stock market, this might manifest as stock prices fluctuating less frequently but showing greater differences between weekly highs and lows.
However, these indicators are less reliable in “noisy” systems, where measurements are not very precise. Many real-world systems are highly noisy, making it difficult to track their behavior accurately.
Are there reliable indicators for real-world systems? To explore this, we examined over 7,000 different methods, aiming to identify one robust enough to perform well even in noisy systems.
Amidst our extensive search, we discovered a few standout methods that excelled at this challenging task. Using these methods, we developed a straightforward new approach for predicting critical points.
We named this method with an appropriately cool acronym: RAD, which stands for “Rescaled AutoDensity.”
We tested our new method on highly detailed recordings of brain activity in mice, focusing on regions of the brain involved in processing visual information.
When a neuron fires, nearby neurons can either transmit the signal further or let it dissipate. Amplified signals have a greater impact, but excessive amplification can push the system past a critical point, potentially leading to a seizure.
RAD Method Reveals Brain Regions Closer to Critical Points and Their Functions
Our RAD method showed that certain brain regions exhibit stronger indicators of being near a critical point than others. Specifically, regions responsible for simpler functions (like recognizing the size and orientation of objects) operate further from a critical point compared to regions handling more complex tasks.
This implies that the brain may have evolved to leverage critical points to enhance its extraordinary computational capabilities.
It’s logical that being far from a critical point (like safe lemmings away from the cliff edge) would lead to highly stable neural activity, which would aid in the efficient and reliable processing of basic visual information.
However, our findings also indicate a benefit to operating near the edge of a critical point. Brain regions in this state might have an extended “memory,” enabling them to perform more complex computations, such as interpreting the overall meaning of an image.
The concept of systems operating near or far from a critical point is relevant in various fields, from finance to medicine. Our research offers a more effective method for understanding these systems and identifying when they might undergo sudden (and often disastrous) changes.
This approach could pave the way for future breakthroughs, such as alerting individuals with epilepsy to imminent seizures or forecasting an impending financial crisis.
Read the original article on: Science Alert
