Surprises in high-dimensional ridgeless least squares interpolation 论文
2022The Annals of Statistics引用 490
Sparse and Compressive Sensing TechniquesImage and Signal Denoising MethodsNumerical methods in inverse problems
摘要
). We recover-in a precise quantitative way-several phenomena that have been observed in large-scale neural networks and kernel machines, including the "double descent" behavior of the prediction risk, and the potential benefits of overparametrization.