notesum.ai

Published at November 4

Breaking the Reclustering Barrier in Centroid-based Deep Clustering

cs.LG

cs.AI

Released Date: November 4, 2024

Authors: Lukas Miklautz¹, Timo Klein¹, Kevin Sidak¹, Collin Leiber², Thomas Lang¹, Andrii Shkabrii¹, Sebastian Tschiatschek¹, Claudia Plant¹

Aff.: ¹Faculty of Computer Science, University of Vienna, Vienna, Austria; ²LMU Munich, Munich, Germany

Arxiv: http://arxiv.org/abs/2411.02275v1

Refer to caption

Methods	CIFAR10			CIFAR100-20
Methods	ACC	NMI	ARI	ACC	NMI	ARI
Pretraining + $k$ -Means	$68.97$	$63.98$	$40.13$	$37.22$	$42.25$	$14.86$
DEC	$88.29$	$80.60$	$77.23$	$50.16$	$51.66$	$\mathbf{35.37}$
DEC + BRB	$\mathbf{90.57}$	$\mathbf{82.57}$	$\mathbf{81.18}$	$\mathbf{50.46}$	$\mathbf{51.72}$	$35.05$
IDEC	88.30	79.50	77.27	52.73	52.79	36.79
IDEC + BRB	$\mathbf{90.72}$	$\mathbf{83.26}$	$\mathbf{81.81}$	$\mathbf{55.43}$	$\mathbf{54.81}$	$\mathbf{38.81}$
DCN	88.55	81.02	78.17	53.27	52.13	37.30
DCN + BRB	$\mathbf{91.23}$	$\mathbf{83.66}$	$\mathbf{82.42}$	$\mathbf{56.92}$	$\mathbf{56.76}$	$\mathbf{41.15}$
SCAN (Gansbeke et al., 2020)	88.3	79.7	77.2	50.7	48.6	33.3
GCC (Zhong et al., 2021)	90.1	-	-	52.3	-	-
SeCu (Qian, 2023)	$93.0$	$86.1$	$85.7$	$55.2$	$55.1$	$39.7$