Double Descent and Overparameterization in Particle Physics Data

Matthias Vigl, Lukas Heinrich

Recently, the benefit of heavily overparameterized models has been observed in machine learning tasks: models with enough capacity to easily cross the \emph{interpolation threshold…