Services and Projects

CANDELA

Estimating distances in the universe with machine learning and historical data The CANDELA project develops innovative methods using Machine Learning and Deep Learning (an advanced type of machine learning based on complex neural networks) to estimate the distance of stars and galaxies. These estimates are based on data collected from astronomical catalogs such as Gaia DR3 and OGLE, which contain observations of variable stars—stars whose brightness changes over time. In particular, the project focuses on two important types of variable stars:

  • Cepheids: stars that pulse regularly, increasing and decreasing in luminosity with a well-defined period. Their luminosity is directly linked to the pulsation period, which allows astronomers to use them as “standard candles” to measure cosmic distances.

  • RR Lyrae: variable stars similar to Cepheids, but of a different and older type, also used to estimate distances in specific regions of the universe.

In parallel, CANDELA analyzes a vast archive of historical astronomical photographs preserved by the Turin Observatory to identify objects of interest and integrate this information into the estimation models.

Financing authority

EU - Next Generation EU EU - Next Generation EU

CANDELA is funded by the European Union (NextGenerationEU) through the NRRP Cascade Calls – Spoke 3, with the objective of developing advanced tools for astrophysics.

Advanced models for cosmic distance

Using machine learning and deep learning models, CANDELA aims to precisely estimate the distance of variable stars such as Cepheids and RR Lyrae, as well as galaxies, by analyzing their light variations over time. The models also include an uncertainty assessment in the estimates, which is essential to ensure scientific accuracy.

Accuracy and scientific insight

The project expects to develop three specific machine learning and deep learning models, analyze approximately 10,000 historical photographic plates, and improve the understanding of uncertainties in astronomical distance estimates.

AI techniques applied to astrophysics

The main innovation is the combined application of machine learning and deep learning techniques, integrated into a single model that also evaluates prediction uncertainty. This approach can have a significant impact on cosmology—for example, by improving the measurement of the Hubble constant, which describes the expansion of the universe.

European research and historical observatories

The project involves various entities and leverages historical archives, such as that of the Astrophysical Observatory of Turin, to integrate data and refine the models.