no code implementations • 9 Jun 2020 • Andrei Boiarov, Oleg Granichin, Olga Granichina
In this paper, we suggest to consider the new multi-task loss function and propose the SPSA-like few-shot learning approach based on the prototypical networks method.