The paper discusses the application of the machine learning library, CatBoost, to determine the masses of radioactive isotopes from the supernova light curve at a later epochs. The synthetic light curve model used for the demonstration is based on the contribution of the five major radioactive decay chains starting with ⁵⁶Ni, ⁵⁷Ni, ⁴⁴Ti, ²²Na, ⁶⁰Co. Separately, we considered sets of random light curves calculated for different isotope masses of both the three dominant chains (⁵⁶Ni, ⁵⁷Ni, ⁴⁴Ti) and all five. It is shown that the masses of dominant isotopes are determined with acceptable accuracy in both cases, even with the standard settings of the machine learning algorithm. In the second case, the accuracy of determining the masses of the other two isotopes (²²Na, ⁶⁰Co) turns out to be unsatisfactory, probably due to their weak contribution to the total light curve.

machine learning, supernovae, light curves

Mathematical modelling in actual problems of science and technics

Urvachev Egor Mikhailovich,  urvachevyegor@gmail.com,  orcid.org/0000-0002-1287-8940,  KIAM RAS