Unlearning-based Neural Interpretations

CL Choi, A Duplessis, S Belongie - arXiv preprint arXiv:2410.08069, 2024 - arxiv.org
Gradient-based interpretations often require an anchor point of comparison to avoid
saturation in computing feature importance. We show that current baselines defined using …

Enhancing Pre-trained Deep Learning Model with Self-Adaptive Reflection

X Wang, M Li, H Yu, C Wang, V Sugumaran… - Cognitive …, 2024 - Springer
In the text mining area, prevalent deep learning models primarily focus on mapping input
features to result of predicted outputs, which exhibit a deficiency in self-dialectical thinking …

Data-Faithful Feature Attribution: Mitigating Unobservable Confounders via Instrumental Variables

Q Sun, H Xia, J Liu - The Thirty-eighth Annual Conference on …, 2024 - openreview.net
The state-of-the-art feature attribution methods often neglect the influence of unobservable
confounders, posing a risk of misinterpretation, especially when it is crucial for the …

Backdoor-based Explainable AI Benchmark for High Fidelity Evaluation of Attribution Methods

P Yang, N Akhtar, J Jiang, A Mian - arXiv preprint arXiv:2405.02344, 2024 - arxiv.org
Attribution methods compute importance scores for input features to explain the output
predictions of deep models. However, accurate assessment of attribution methods is …

QUCE: The Minimisation and Quantification of Path-Based Uncertainty for Generative Counterfactual Explanations

J Duell, M Seisenberger, H Fu, X Fan - arXiv preprint arXiv:2402.17516, 2024 - arxiv.org
Deep Neural Networks (DNNs) stand out as one of the most prominent approaches within
the Machine Learning (ML) domain. The efficacy of DNNs has surged alongside recent …

Denoising Diffusion Path: Attribution Noise Reduction with An Auxiliary Diffusion Model

Y Lei, Z Li, J Zhang, H Shan - The Thirty-eighth Annual Conference on … - openreview.net
The explainability of deep neural networks (DNNs) is critical for trust and reliability in AI
systems. Path-based attribution methods, such as integrated gradients (IG), aim to explain …

[引用][C] Explainable Artificial Intelligence for Medical Science

J DUELL - 2024