Reconstructing rotation curves with artificial neural networks

Galactic rotation curves have been served as indispensable tools for determining the distribution of mass within galaxies. Despite several advances in precision observations, some discrepancies still persist between the inferred matter distribution from luminosity and observed rotation velocities, often attributed to the presence of dark matter. Traditional parametric models, while insightful, struggle with the complexity of galaxies with prominent bulges or non-circular motions, but in contrast, non-parametric methods offer a promising alternative, adapting to the intricate nature of rotation curves without any prior assumptions. In this work, we employ artificial neural networks to reconstruct rotation curves of 17 spiral galaxies from high-quality data, demonstrating the efficacy of the non-parametric approaches in characterizing galactic dynamics. Our findings underscore the importance of employing diverse methodological approaches to comprehensively understand galactic dynamics in modern astrophysics research. Moreover, the non-parametric reconstruction approach with neural networks presents a promising avenue for further investigations, capable of generating interpolations based on the intrinsic patterns of the data.