Geodesic
Explaining Deep Residual Networks Predictions with Symplectic Adjoint Method
Computer Science and Information Systems, Tome 20 (2023) no. 4 Cet article a éte moissonné depuis la source Computer Science and Information Systems website

Voir la notice de l'article

Understanding deep residual networks (ResNets) decisions are receiving much attention as a way to ensure their security and reliability. Recent research, however, lacks theoretical analysis to guarantee the faithfulness of explanations and could produce an unreliable explanation. In order to explain ResNets predictions, we suggest a provably faithful explanation for ResNet using a surrogate explainable model, a neural ordinary differential equation network (Neural ODE). First, ResNets are proved to converge to a Neural ODE and the Neural ODE is regarded as a surrogate model to explain the decision-making attribution of the ResNets. And then the decision feature and the explanation map of inputs belonging to the target class for Neural ODE are generated via the symplectic adjoint method. Finally, we prove that the explanations of Neural ODE can be sufficiently approximate to ResNet. Experiments show that the proposed explanation method has higher faithfulness with lower computational cost than other explanation approaches and it is effective for troubleshooting and optimizing a model by the explanation.
Keywords: Deep residual networks, Explanation, Neural ODE, Symplectic adjoint method 
@article{CSIS_2023_20_4_a8,
     author = {Xia Lei and Jia-Jiang Lin and Xiong-Lin Luo and Yongkai Fan},
     title = {Explaining {Deep} {Residual} {Networks} {Predictions} with {Symplectic} {Adjoint} {Method}},
     journal = {Computer Science and Information Systems},
     year = {2023},
     volume = {20},
     number = {4},
     url = {http://geodesic.mathdoc.fr/item/CSIS_2023_20_4_a8/}
}
TY  - JOUR
AU  - Xia Lei
AU  - Jia-Jiang Lin
AU  - Xiong-Lin Luo
AU  - Yongkai Fan
TI  - Explaining Deep Residual Networks Predictions with Symplectic Adjoint Method
JO  - Computer Science and Information Systems
PY  - 2023
VL  - 20
IS  - 4
UR  - http://geodesic.mathdoc.fr/item/CSIS_2023_20_4_a8/
ID  - CSIS_2023_20_4_a8
ER  - 
%0 Journal Article
%A Xia Lei
%A Jia-Jiang Lin
%A Xiong-Lin Luo
%A Yongkai Fan
%T Explaining Deep Residual Networks Predictions with Symplectic Adjoint Method
%J Computer Science and Information Systems
%D 2023
%V 20
%N 4
%U http://geodesic.mathdoc.fr/item/CSIS_2023_20_4_a8/
%F CSIS_2023_20_4_a8
Xia Lei; Jia-Jiang Lin; Xiong-Lin Luo; Yongkai Fan. Explaining Deep Residual Networks Predictions with Symplectic Adjoint Method. Computer Science and Information Systems, Tome 20 (2023) no. 4. http://geodesic.mathdoc.fr/item/CSIS_2023_20_4_a8/