Optimal learning rates for Kernel Conjugate Gradient regression

Gilles Blanchard, Nicole Kraemer

We prove rates of convergence in the statistical sense for kernel-based least squares regression using a conjugate gradient algorithm, where regularization against overfitting is o…