Full Text:
<49>
CLC number:
On-line Access: 2025-04-17
Received: 2024-11-25
Revision Accepted: 2025-01-26
Crosschecked: 0000-00-00
Cited: 0
Clicked: 72
A survey towards the first principle of explaining DNNs:interactions explain the learning dynamics
| |||||||||||||
Huilin ZHOU, Qihan REN, Junpeng ZHANG, Quanshi ZHANG‡. A survey towards the first principle of explaining DNNs:interactions explain the learning dynamics[J]. Frontiers of Information Technology & Electronic Engineering, 1998, -1(-1): .
@article{title="A survey towards the first principle of explaining DNNs:interactions explain the learning dynamics",
author="Huilin ZHOU, Qihan REN, Junpeng ZHANG, Quanshi ZHANG‡",
journal="Frontiers of Information Technology & Electronic Engineering",
volume="-1",
number="-1",
pages="",
year="1998",
publisher="Zhejiang University Press & Springer",
doi="10.1631/FITEE.2401025"
}
%0 Journal Article
%T A survey towards the first principle of explaining DNNs:interactions explain the learning dynamics
%A Huilin ZHOU
%A Qihan REN
%A Junpeng ZHANG
%A Quanshi ZHANG‡
%J Journal of Zhejiang University SCIENCE C
%V -1
%N -1
%P
%@ 2095-9184
%D 1998
%I Zhejiang University Press & Springer
%DOI 10.1631/FITEE.2401025
TY - JOUR
T1 - A survey towards the first principle of explaining DNNs:interactions explain the learning dynamics
A1 - Huilin ZHOU
A1 - Qihan REN
A1 - Junpeng ZHANG
A1 - Quanshi ZHANG‡
J0 - Journal of Zhejiang University Science C
VL - -1
IS - -1
SP -
EP -
%@ 2095-9184
Y1 - 1998
PB - Zhejiang University Press & Springer
ER -
DOI - 10.1631/FITEE.2401025
Abstract: Most explanation methods are designed in an empirical manner, so whether there exists a first-principal explanation of deep neural networks (DNNs) becomes the next core scientific problem in explainable artificial intelligence (XAI). Although it is still an open problem, in this paper, we discuss whether the interaction-based explanation can serve as the first-principal explanation of a DNN. The strong explanatory power of interaction theory comes from the following aspects: (1) it establishes a new axiomatic system to quantify the decision-making logic of a DNN into a set of symbolic interaction concepts; (2) it simultaneously explains various deep learning phenomena, such as generalization power, adversarial sensitivity, representation bottleneck, and learning dynamics; (3) it provides mathematical tools that uniformly explain the mechanisms of various empirical attribution methods and empirical adversarial-transferability-boosting methods; and (4) it explains the extremely complex learning dynamics of a DNN by analyzing the two-phase dynamics of interaction complexity, which further reveals the internal mechanism of why and how the generalization power/adversarial sensitivity of a DNN changes during the learning process.
Open peer comments: Debate/Discuss/Question/Opinion
<1>