= Early stopping of Landweber iteration
The <Landweber spectral filter> progressively admits smaller singular-value components. Its approximation bias tends to zero on exact admissible data, but its noise amplification grows. With $\|A\|\leq1$, a sufficient <a priori regularization parameter choice> is $n(\delta)\to\infty$ and $\sqrt{n(\delta)}\,\delta\to0$. Taking regularization parameter $\alpha=1/n$ expresses this as $\alpha\to0$ and $\delta/\sqrt\alpha\to0$; the <noise-bias decomposition for linear regularization> proves convergence.
Back to article page
