On the Universality of the Double Descent Peak in Ridgeless Regression

David Holzmüller

Keywords: Double Descent, Interpolation Peak, Linear Regression, Random Features, Random Weights Neural Networks

Abstract: We prove a non-asymptotic distribution-independent lower bound for the expected mean squared generalization error caused by label noise in ridgeless linear regression. Our lower bound generalizes a similar known result to the overparameterized (interpolating) regime. In contrast to most previous works, our analysis applies to a broad class of input distributions with almost surely full-rank feature matrices, which allows us to cover various types of deterministic or random feature maps. Our lower bound is asymptotically sharp and implies that in the presence of label noise, ridgeless linear regression does not perform well around the interpolation threshold for any of these feature maps. We analyze the imposed assumptions in detail and provide a theory for analytic (random) feature maps. Using this theory, we can show that our assumptions are satisfied for input distributions with a (Lebesgue) density and feature maps given by random deep neural networks with analytic activation functions like sigmoid, tanh, softplus or GELU. As further examples, we show that feature maps from random Fourier features and polynomial kernels also satisfy our assumptions. We complement our theory with further experimental and analytic results.

One-sentence Summary: We prove a distribution-independent lower bound for the generalization error of ridgeless (random) features regression under weak assumptions, showing universal sensitivity to label noise around the interpolation threshold.

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics

Code: [ dholzmueller/universal_double_descent](https://github.com/dholzmueller/universal_double_descent)

Community Implementations: [ 1 code implementation](https://www.catalyzex.com/paper/arxiv:2010.01851/code)