notesum.ai

Published at November 15

Federated Domain Generalization via Prompt Learning and Aggregation

cs.AI

cs.CV

cs.LG

Released Date: November 15, 2024

Authors: Shuai Gong, Chaoran Cui, Chunyun Zhang, Wenna Wang, Xiushan Nie, Lei Zhu

Arxiv: http://arxiv.org/abs/2411.10063v1

Refer to caption

Methods	PACS
Methods	Art	Cartoon	Photo	Sketch	Avg.
CL-based DG methods
SWAD	93.23	85.93	99.18	82.03	90.44
HCVP	93.17	86.89	99.33	81.30	90.17
Doprompt	95.00	86.35	99.63	78.20	89.91
FL-based DG methods
FedAvg	82.67	65.27	97.42	65.18	77.64
FedProx	83.89	68.39	97.43	64.21	78.48
Conventional FedDG methods
FedSR	88.19	67.45	95.74	65.92	79.33
FedADG	82.93	65.42	98.09	65.36	77.95
CCST	87.02	74.57	98.29	65.84	81.43
ELCFS	86.77	73.21	98.14	65.16	80.82
ELCFS+GA	87.68	75.19	97.56	65.86	81.57
PEFT-based methods
FedCLIP	96.19	97.91	99.76	85.85	94.93
PromptFL	96.34	98.46	99.58	92.19	96.64
FedAPT	97.15	99.12	99.69	92.34	97.08
FedPR	98.10	99.02	99.88	91.11	97.03
FedMaPLe	98.44	99.02	99.94	90.40	96.95
PLAN (Ours)	98.58	99.14	99.82	92.08	97.40