notesum.ai

Published at October 18

Towards Unsupervised Validation of Anomaly-Detection Models

cs.CL

cs.AI

Released Date: October 18, 2024

Authors: Lihi Idan¹

Aff.: ¹Texas A&M University

Arxiv: https://arxiv.org/abs/2410.14579v1

Refer to caption

Dataset	Best ensemble	Worst ensemble	AS	RSPS	% Improvement w.r.t RSPS	[25],[23],[35],[34] (PR AUC)
smtp [29]	.39, .55	.02, .06	.19, .35	.11, .2	254, 175	.2,.001,.09,.04
mnist [16]	.41, .42	.15, .17	.3, .31	.26, .3	57, 40	.19,.24,.34,.27
backdoor [22]	.48, .45	.04, .06	.28, .29	.2, .25	140, 80	.02,.2,.13,.27
gamma [7]	.48, .43	.35, .38	.39, .37	.37, .35	29, 23	.35,.33,.44,.43
fraud [24]	.29, .34	.02, .06	.12, .19	.12, .18	141, 88	.08,.001,.27,.15
campaign [24]	.34, .38	.13, .13	.25, .3	.23, .29	48, 31	.21,.15,.24,.22
satellite [28]	.6, .55	.25, .3	.49, .46	.44, .44	36, 25	.5,.23,.45,.36
pendigits [2]	.28, .35	.03, .03	.16, .2	.12, .18	133, 94	.21,.05,.08,.08
shuttle [28]	.92, .81	.13, .15	.47, .48	.36, .47	155, 72	.1,.08,.51,.65