Shaping double descent in quantum extreme learning through model design
This work analytically characterizes the entire risk curve through two quantities, namely the condition number of the feature population covariance and a spectral measure of correlation strength, focusing on quantum extreme learning machines (QELMs).