Old Ideas, New Discoveries: Armenian Statistical Methods Find New Applications in Astronomy and Cosmology

YEREVAN  – Scientific ideas sometimes have to wait decades for technology to catch up. Statistical algorithms developed at the Yerevan Physics Institute more than 35 years ago for high-energy particle physics are now finding new applications in astronomy and cosmology, revealing hidden structures in stellar populations and the large-scale distribution of galaxies.

This is evidenced by the work completed by the Cosmic Ray Division (CRD) of the Yerevan Physics Institute and CRD’s head, Professor Ashot Chilingarian, together with CRD’s staff and colleagues.

Professor Ashot Chilingarian has revived and modernized two systems, the ANI (Advanced Nonparametric Inference) package and the Monte Carlo Statistical Inference (MCSI) methodology, that he originally developed in the late 1980s and early 1990s. At the time, available computers were simply too slow to exploit their full potential. Today’s powerful processors and the enormous astronomical databases collected by modern sky surveys have changed the situation completely.

During the past two years these classical statistical analysis methods have been adapted to different astronomical problems. In one study, the CRD’s algorithm identified the combinations of chemical elements that distinguish different generations of stars in globular clusters, providing a transparent multivariate alternative to conventional AI machine-learning techniques. In another, CRD’s statistical analyses algorithm analyzed the three-dimensional distribution of thousands of galaxies from the Sloan Digital Sky Survey, revealing the local geometry of the “cosmic web”—compact galaxy groups, with elongated filaments, sheets, and vast cosmic voids. A third study demonstrated that the same mathematical framework reliably reconstructs astronomical structures in realistic simulations, showing that the approach is applicable over an enormous range of spatial scales—from individual stellar systems to the large-scale structure of the Universe.

Although developed long before the current boom in artificial intelligence, ANI follows a philosophy that is highly relevant today. Rather than treating data as a “black box,” it searches directly for statistically significant structures hidden in multidimensional space. CRD’s algorithms identify combinations of variables that carry physical meaning while remaining transparent and interpretable—an increasingly important requirement in modern scientific data analysis.

The revival of CRD’s ANI analysis package illustrates that scientific progress does not always require entirely new ideas. Sometimes, advances in computing make it possible to realize the full potential of concepts that were developed decades earlier. Methods originally designed to analyze particle interactions at accelerator experiments are now helping astronomers investigate stellar evolution, galaxy formation, and the architecture of the Universe.

Beyond astronomy and cosmology, the same statistical framework is applicable to genomics, environmental sciences, medicine, finance, and other fields where important information is hidden within large multidimensional datasets. This work demonstrates that carefully designed statistical inference remains a powerful complement to modern artificial intelligence, particularly when physical interpretation is as important as predictive performance.