Mutual information-entropy plane: a new quantifier space for time series analysis

In the study of complex systems, from the flickering of a stock market to the firing of neurons in a brain, scientists often look for ways to measure how unpredictable a system is and how much it shares with another. Two powerful tools have long helped researchers navigate this terrain: entropy and mutual information. Entropy is a measure of disorder or randomness; a perfectly predictable clock has low entropy, while a system of pure static noise has high entropy. Mutual information, on the other hand, measures how much two different systems tell us about each other. If knowing the state of one system allows you to guess the state of another with high accuracy, they share a lot of mutual information. For decades, scientists have used these tools separately, plotting them on maps to distinguish between orderly, chaotic, and random behaviors. However, a significant gap remained: while these tools could describe a single system's chaos or the link between two systems, they struggled to show both the internal uncertainty of a system and its connection to another at the same time, nor could they easily reveal which system was driving the other.

A team of researchers from Argentina has now proposed a new way to visualize these relationships, creating a two-dimensional map that combines these concepts into a single, intuitive picture. By taking the standard measure of disorder and pairing it with a measure of shared information, they constructed a plane where every point tells a complete story about a pair of interacting time series. This new approach, which they call the mutual information-entropy plane, allows scientists to see not just how chaotic a system is or how closely it is linked to a partner, but also to determine the direction of influence between them. The researchers tested this method on three distinct types of behavior: regular, predictable motion; chaotic, deterministic motion that looks random but follows strict rules; and purely random noise. They found that this new map could clearly separate these different behaviors, even when they shared the same amount of information, solving a problem that previous methods could not.

The core of this new method lies in how the researchers arranged their data. They took the measure of a system's internal disorder and placed it on one axis, while the measure of how much information it shared with a partner went on the other. When they plotted the results, a clear geometric pattern emerged. The researchers discovered that the distance of a data point from a specific diagonal line on the map reveals something profound: the degree to which one system remains independent of the other. If a point sits directly on the line, the two systems are so tightly coupled that knowing one completely explains the other. If the point sits far below the line, it means that even though the systems might be linked, a significant portion of one system's behavior remains a mystery when looking at the other. This vertical distance acts as a direct measure of "informational independence," a concept that was difficult to visualize before.

What makes this discovery particularly powerful is its ability to reveal directionality without complex calculations. In many scientific fields, determining whether system A causes changes in system B, or vice versa, requires heavy computational tools. The researchers showed that by plotting both systems on this same map, the geometry itself reveals the flow of influence. If system A drives system B, the point representing A will sit closer to the diagonal line than the point representing B. This asymmetry creates a simple visual arrow, pointing from the driver to the driven, allowing researchers to see the direction of information flow just by looking at the relative positions of the points. It is a way of seeing who is leading and who is following in a dance of data, without needing to calculate the steps of the dance itself.

To prove the utility of their new map, the researchers simulated three classic scenarios. First, they looked at regular, predictable motion, like a pendulum swinging. Next, they examined chaotic systems, which are deterministic but appear random, such as the famous logistic map used to model population growth. Finally, they studied pure white noise, which is completely random. They then created a second system for each case that was a mixture of the original and some random noise, controlled by a coupling factor. As they adjusted this factor, they watched how the points moved across their new map. They found that while the amount of shared information increased steadily as the systems became more coupled, the position of the points on the map remained distinct for each type of dynamics. The regular, chaotic, and random systems occupied different regions of the plane, allowing them to be distinguished clearly. This was a crucial finding because previous methods often struggled to tell chaotic systems apart from random noise when they shared similar levels of information.

The researchers also demonstrated that this method works for real-world applications where direction matters. By plotting the entropy of the driving system against the shared information, and then plotting the entropy of the resulting system against the same shared information, they could see a clear split. The point for the driving system stayed closer to the line of perfect dependence, while the point for the resulting system drifted further away. This visual gap confirmed that the information was flowing in one direction, from the driver to the receiver. The researchers noted that this approach offers a more direct and geometric way to understand these relationships compared to older, more computationally expensive techniques.

However, the authors are careful to note the limitations of their new tool. Like any method that relies on statistical patterns, it is sensitive to how the data is prepared. The way the time series are broken down, the size of the windows used for analysis, and the amount of data available can all shift the position of the points on the map. For very short or noisy data sets, the map might lose some of its sharpness, making it harder to distinguish between subtle differences. In these cases, the researchers suggest treating the results as general trends rather than precise measurements. Despite these constraints, the new plane provides a robust framework for comparing complex dynamics. It offers a way to simultaneously characterize the internal uncertainty of a system and its relationship with another, turning abstract mathematical concepts into a visual landscape that researchers can navigate with clarity.

This work represents a significant step forward in the analysis of complex systems, offering a unified view of uncertainty and connection. By combining the measures of disorder and shared information into a single, two-dimensional space, the researchers have provided a tool that is both intuitive and powerful. It allows scientists to see not only how chaotic or random a system is, but also how it relates to others and in which direction influence flows. Whether applied to financial markets, brain activity, or physical systems, this new map promises to help researchers untangle the intricate web of cause and effect that defines our complex world, turning invisible patterns of information into a clear, visual story.