The 35th annual conference on Astronomical Data Analysis Software & Systems 2025 in Görlitz, Germany

Accurate classification of astronomical light curves is important for interpreting the large datasets produced by modern surveys. However, most existing feature sets overlook the nonlinear dynamics inherent to the astrophysical systems that generate these signals. In this work, we explore features derived from nonlinear time-series analysis and assess their utility for light-curve classification. Using the PLAsTiCC benchmark dataset, we find that nonlinear features can perform competitively on their own and may improve classification when combined with standard time-series features. These preliminary results suggest that nonlinear descriptors capture complementary aspects of astrophysical variability and could provide value in future large-scale surveys.