How to select interpretable circular coordinates from data

Select interpretable circular coordinates with CIRCOL, a method based on persistent cohomology. Ideal for data scientists.

viernes, 10 de julio de 2026 • 6 min read • Q2BSTUDIO Team

Efficient selection of interpretable circular coordinates

In today's world, data generated by sensors, molecular simulations, or neural recordings contains a wealth of information that is often hidden under complex structures. One of the most promising techniques for unveiling these structures is the use of circular coordinates obtained from persistent cohomology. However, these coordinates are usually abstract: a detected circle does not tell us which angle, phase, torsion or decoder explains it. The need to interpret these results has led to the development of methods for selecting circular coordinates that are scientifically significant, an approach that combines advanced mathematics, artificial intelligence, and real business needs.

The key is to move from a purely topological detection – "there is a loop in the data" – to a concrete interpretation: "that loop corresponds to the angle of a rotor in a motor, or to the phase of a neural oscillation". This requires a selection process from a user-provided dictionary of candidates, each represented by the cohomology class of its angular shape. In the continuous scope, selecting the set of least-energy coordinates that explains the relevant subspace becomes a minimum weight base problem in a vector matroid. For discrete point clouds, techniques such as the CIRCOL method demonstrate that it is possible to consistently estimate the internal product of smooth 1-forms even under non-uniform sampling, thus allowing the construction of projection matrices that guide selection and diagnose unexplained persistent classes.

From a professional perspective, this capability has direct applications in sectors such as engineering, computational biology, and neuroscience. For example, in molecular simulations, identifying the relevant angular degrees of freedom can help to understand conformational transitions of proteins. In the analysis of head direction cell recordings in rodents, circular coordinates reveal the cognitive map of the animal. For companies that work with big data, having tools that transform abstract topology into actionable insights is a competitive differentiator.

This is where the need for AI for businesses and custom software solutions comes into play. A platform that implements these algorithms requires careful development, from the integration of computational topology libraries to the interactive visualization of the results. At Q2BSTUDIO, as a software and technology development company, we understand that every business has unique problems. That's why we offer bespoke applications that not only run complex analytics, but present them in ways that are understandable to data science teams.

In addition, the scalability of these processes is supported by AWS and Azure cloud services. Hosting persistent cohomology calculations in the cloud allows you to handle massive datasets without compromising performance. Combined with business intelligence services such as power BI, it is possible to generate dashboards that monitor in real time the topological structures emerging in the data. For example, in an industrial environment, a Power BI dashboard could alert engineers to the appearance of a new cycle in a machine's vibration signals, indicating a potential incipient failure.

Cybersecurity also benefits from this approach. Detecting unusual patterns in network traffic can be modeled as the appearance of unexpected circular coordinates in feature space. With the right AI agents, it is possible to automate the response to these events. Q2BSTUDIO integrates these capabilities into customized solutions, ensuring that artificial intelligence does not remain in a black box, but offers interpretable explanations.

From a technical point of view, the method described in the literature – such as the one that inspires this article – proposes an approach based on energy minimisation. In simple terms, it is about finding a set of circular coordinates that are "orthogonal" to each other and that represent the variability of the data in the most efficient way. This is reminiscent of principal component analysis, but in a circular context. The crucial difference is that circular coordinates preserve the topology of space, something that linear PCA cannot capture when data lives on a torus or a sphere. For this reason, these techniques are especially useful in domains where the variables are angular or periodic, such as in robotics (joints), astronomy (orbital phases) or economics (seasonal cycles).

A fascinating aspect of recent research is the demonstration that the internal product of cochains (discrete versions of differential shapes) can consistently estimate the L² internal product of soft 1-forms, even when the sample is not uniform. This solves an important practical problem: in many real applications, the dots are not evenly distributed. For example, in a molecular simulation, certain conformations may be more populated than others. The method must be robust to these density variations. The resulting projection matrix not only helps to select the dictionary coordinates with low energy, but also diagnoses those persistent classes that are not explained by the proposed candidates, indicating that perhaps new variables need to be added to the dictionary.

For companies looking to implement these techniques, the challenge is not only mathematical, but also one of integration. You need a platform that manages the entire pipeline: from data acquisition, through the calculation of persistent cohomology, to the selection of interpretable coordinates and visualization. This is where Q2BSTUDIO adds value. Our team of custom software experts can build systems that automate this workflow, using containers and cloud orchestration so data scientists focus on interpretation, not infrastructure.

In addition, the combination with AI agents allows the system to learn to suggest candidates from the dictionary automatically. For example, if a set of sensor data shows a persistent circle, an intelligent agent could propose the phase of the first harmonic of the signal as a candidate, based on previous analyses. This level of automation reduces analysis time and democratizes access to advanced computational topology techniques.

In cybersecurity, imagine an intrusion detection system that, instead of fixed thresholds, uses circular coordinates to model the normal behavior of the network. Any deviation that generates a new topological cycle could be a sign of attack. By combining this with AWS and Azure cloud services, it is possible to scale the analysis to terabytes of real-time network traffic. And all this is reflected in a power bi dashboard accessible to the security team.

Business intelligence benefits from these techniques by being able to identify cyclical patterns in sales, demand or prices of raw materials. Interpretable circular coordinates allow, for example, to distinguish between a weekly cycle and a monthly cycle, and to understand how they interact. With bespoke applications developed by Q2BSTUDIO, companies can integrate this analysis directly into their strategic planning processes.

In short, the selection of interpretable circular coordinates is much more than an academic exercise: it is a gateway to deeper data analysis, connecting topology to business practice. Whether it's in molecular simulations, neural recordings, or industrial monitoring, the ability to extract meaning from hidden loops in data makes all the difference. Companies like Q2BSTUDIO are at the forefront of offering solutions that transform these advanced concepts into functional and scalable tools. If your organization handles data with periodic structure or seeks to better understand nonlinear relationships in its metrics, consider how the combination of enterprise AI, custom software, and cloud services can unlock that potential. The future of data analytics is no longer just about predicting; it is interpreting.

OUR SERVICES

How we can help you

Do you have a project in mind?

Tell us your vision and we'll turn it into a software solution. Whatever the scope, we make your idea real.