A PAC-Bayesian Generalization Bound for Equivariant Networks
被引:0
作者:
Behboodi, Arash
论文数: 0引用数: 0
h-index: 0
机构:
Qualcomm AI Res, Amsterdam, NetherlandsQualcomm AI Res, Amsterdam, Netherlands
Behboodi, Arash
[1
]
Cesa, Gabriele
论文数: 0引用数: 0
h-index: 0
机构:
Qualcomm AI Res, Amsterdam, Netherlands
Univ Amsterdam, AMLab, Amsterdam, NetherlandsQualcomm AI Res, Amsterdam, Netherlands
Cesa, Gabriele
[1
,2
]
Cohen, Taco
论文数: 0引用数: 0
h-index: 0
机构:
Qualcomm AI Res, Amsterdam, NetherlandsQualcomm AI Res, Amsterdam, Netherlands
Cohen, Taco
[1
]
机构:
[1] Qualcomm AI Res, Amsterdam, Netherlands
[2] Univ Amsterdam, AMLab, Amsterdam, Netherlands
来源:
ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 35 (NEURIPS 2022)
|
2022年
关键词:
GROUP INVARIANT;
D O I:
暂无
中图分类号:
TP18 [人工智能理论];
学科分类号:
081104 ;
0812 ;
0835 ;
1405 ;
摘要:
Equivariant networks capture the inductive bias about the symmetry of the learning task by building those symmetries into the model. In this paper, we study how equivariance relates to generalization error utilizing PAC Bayesian analysis for equivariant networks, where the transformation laws of feature spaces are determined by group representations. By using perturbation analysis of equivariant networks in Fourier domain for each layer, we derive norm-based PAC-Bayesian generalization bounds. The bound characterizes the impact of group size, and multiplicity and degree of irreducible representations on the generalization error and thereby provide a guideline for selecting them. In general, the bound indicates that using larger group size in the model improves the generalization error substantiated by extensive numerical experiments.