notesum.ai

Published at December 3

Hanke-Raus heuristic rule for iteratively regularized stochastic gradient descent

math.NA

cs.NA

math.AP

math.CA

math.FA

Released Date: December 3, 2024

Authors: Harshit Bajpai¹, Gaurav Mittal¹, Ankik Kumar Giri¹

Aff.: ¹University

Arxiv: http://arxiv.org/pdf/2412.02397v1

Refer to caption

$\delta_{\text{rel}}$			$a$ in Rule 1
	$1\text{e}1$		$1\text{e}2$		$1\text{e}3$		$1\text{e}3$
	$k_{*}$	$E_{k_{*}}^{\delta}$	$k_{*}$	$E_{k_{*}}^{\delta}$	$k_{*}$	$E_{k_{*}}^{\delta}$	$k_{*}$	$E_{k_{*}}^{\delta}$
$1\text{e}^{-1}$	18	$3.550\text{e}^{-2}$	32	$1.565\text{e}^{-2}$	49	$1.441\text{e}^{-2}$	249	$4.213\text{e}^{-3}$
$1\text{e}^{-2}$	133	$4.817\text{e}^{-3}$	80	$8.626\text{e}^{-3}$	506	$1.300\text{e}^{-3}$	565	$1.664\text{e}^{-3}$
$1\text{e}^{-3}$	172	$2.293\text{e}^{-3}$	260	$4.982\text{e}^{-3}$	2044	$5.845\text{e}^{-4}$	3404	$4.982\text{e}^{-4}$
$1\text{e}^{-4}$	3304	$5.341\text{e}^{-4}$	3167	$5.134\text{e}^{-4}$	867	$6.111\text{e}^{-4}$	4874	$3.185\text{e}^{-4}$