notesum.ai

Published at November 2

FEED: Fairness-Enhanced Meta-Learning for Domain Generalization

cs.LG

cs.AI

Released Date: November 2, 2024

Authors: Kai Jiang¹, Chen Zhao², Haoliang Wang¹, Feng Chen¹

Aff.: ¹Department of Computer Science, The University of Texas at Dallas, Richardson, Texas, USA; ²Department of Computer Science, Baylor University, Waco, Texas, USA

Arxiv: http://arxiv.org/abs/2411.01316v1

	Accuracy $\uparrow$ / $\Delta$ DP $\downarrow$ / $\Delta$ EOPP $\downarrow$ / $\Delta$ EO $\downarrow$
Methods	$d_{0}$	$d_{1}$	$d_{2}$	Avg
ERM [12]	89.69 / 0.133 / 0.005 / 0.007	86.92 / 0.049 / 0.005 / 0.017	87.18 / 0.050 / 0.004 / 0.007	87.93 / 0.077 / 0.004 / 0.011
IRM [2]	67.05 / 0.067 / 0.015 / 0.018	65.80 / 0.044 / 0.009 / 0.015	71.01 / 0.040 / 0.002 / 0.012	67.95 / 0.050 / 0.009 / 0.015
GroupDRO [13]	89.20 / 0.138 / 0.001 / 0.026	66.63 / 0.048 / 0.004 / 0.011	85.99 / 0.048 / 0.003 / 0.002	80.61 / 0.078 / 0.002 / 0.013
Mixup [14]	90.00 / 0.130 / 0.001 / 0.004	86.06 / 0.050 / 0.005 / 0.020	86.70 / 0.049 / 0.002 / 0.007	87.58 / 0.076 / 0.002 / 0.010
DDG [16]	83.74 / 0.093 / 0.032 / 0.067	88.26 / 0.056 / 0.016 / 0.034	89.95 / 0.043 / 0.004 / 0.003	87.32 / 0.064 / 0.018 / 0.035
MBDG [17]	85.70 / 0.136 / 0.029 / 0.024	89.90 / 0.063 / 0.025 / 0.035	87.49 / 0.036 / 0.001 / 0.006	87.70 / 0.079 / 0.019 / 0.022
DDG-FC	86.46 / 0.108 / 0.038 / 0.046	89.32 / 0.067 / 0.030 / 0.038	88.04 / 0.058 / 0.017 / 0.012	87.94 / 0.077 / 0.028 / 0.032
MBDG-FC	92.12 / 0.057 / 0.032 / 0.154	70.72 / 0.061 / 0.001 / 0.002	85.56 / 0.054 / 0.001 / 0.008	82.80 / 0.057 / 0.011 / 0.055
EIIL [19]	71.56 / 0.064 / 0.040 / 0.065	68.96 / 0.049 / 0.009 / 0.006	72.20 / 0.042 / 0.001 / 0.001	70.91 / 0.052 / 0.017 / 0.024
FarconVAE [20]	84.80 / 0.175 / 0.001 / 0.011	72.60 / 0.048 / 0.002 / 0.012	74.50 / 0.071 / 0.004 / 0.012	77.30 / 0.098 / 0.002 / 0.012
FEDORA [18]	87.40 / 0.139 / 0.001 / 0.010	89.50 / 0.020 / 0.002 / 0.008	90.00 / 0.030 / 0.002 / 0.007	88.97 / 0.063 / 0.001 / 0.008
FEED (Ours)	83.96 / 0.060 / 0.001 / 0.008	91.36 / 0.033 / 0.001 / 0.009	92.47 / 0.038 / 0.001 / 0.002	89.26 / 0.044 / 0.001 / 0.006