On the Inconsistency of Kernel Ridgeless Regression in Fixed Dimensions
This challenges a key assumption in machine learning theory about overfitting, showing it does not hold for a common class of predictors, which is incremental but clarifies limitations.
The paper demonstrates that kernel ridgeless regression with translation-invariant kernels does not exhibit benign overfitting in fixed dimensions, as the predictor fails to converge to the ground truth with increasing sample size, regardless of bandwidth selection.
``Benign overfitting'', the ability of certain algorithms to interpolate noisy training data and yet perform well out-of-sample, has been a topic of considerable recent interest. We show, using a fixed design setup, that an important class of predictors, kernel machines with translation-invariant kernels, does not exhibit benign overfitting in fixed dimensions. In particular, the estimated predictor does not converge to the ground truth with increasing sample size, for any non-zero regression function and any (even adaptive) bandwidth selection. To prove these results, we give exact expressions for the generalization error, and its decomposition in terms of an approximation error and an estimation error that elicits a trade-off based on the selection of the kernel bandwidth. Our results apply to commonly used translation-invariant kernels such as Gaussian, Laplace, and Cauchy.