analiza aiqbench
This research introduces VI-EDL, a mathematically grounded framework that addresses critical limitations in Evidential Deep Learning by preventing overconfident predictions through variational inference. By deriving an Evidence Lower Bound and establishing rigorous generalization guarantees, the method significantly improves uncertainty quantification for out-of-distribution detection and noise handling across visual and medical datasets. The work advances machine learning theory with practical implications for safety-critical applications like autonomous driving where reliable confidence metrics are essential.
Variational Inference for Evidential Deep Learning
arXiv:2605.26477v3 Announce Type: replace Abstract: While Deep Neural Networks (DNNs) achieve remarkable performance, their tendency to produce overconfident predictions. Evidential Deep Learning (EDL) mitigates this by formulating predictions as a Dirichlet distribution over class probabilities to explicitly quantify epistemic uncertainty. However, we found that the conventional EDL suffers from two fundamental limitations: a Kullback-Leibler (KL) penalty that only suppresses the evidence of negative classes, producing excessively high evidence therefore decreasing the model's ability to quantify uncertainty, and an absence in theoretical guarantee of setting Dirichlet parameter $\alpha=e+1$. In this paper, we propose a mathematically principled framework, Variational Inference Evidential Deep Learning (VI-EDL). By reformulating evidential learning through the lens of variational inference, we derive an Evidence Lower Bound (ELBO), which prevents the evidence from growing excessively. Theoretically, we rigorously establish a generalization bound and reveal how the predicted uncertainty, feature and network complexity affect this bound, and why setting $\boldsymbol{\alpha} = \mathbf{e} + \mathbf{1}$ can minimize it. Extensive experiments on standard visual and medical datasets demonstrate that VI-EDL achieves state-of-the-art performance, showing excellent performance in out-of-distribution detection, noise detection and autonomous driving scenario. The code is available in https://github.com/seutjw/VI-EDL.
Abstract:While Deep Neural Networks (DNNs) achieve remarkable performance, their tendency to produce overconfident predictions. Evidential Deep Learning (EDL) mitigates this by formulating predictions as a Dirichlet distribution over class probabilities to explicitly quantify epistemic uncertainty. However, we found that the conventional EDL suffers from two fundamental limitations: a Kullback-Leibler (KL) penalty that only suppresses the evidence of negative classes, producing excessively high evidence therefore decreasing the model's ability to quantify uncertainty, and an absence in theoretical guarantee of setting Dirichlet parameter $\alpha=e+1$. In this paper, we propose a mathematically principled framework, Variational Inference Evidential Deep Learning (VI-EDL). By reformulating evidential learning through the lens of variational inference, we derive an Evidence Lower Bound (ELBO), which prevents the evidence from growing excessively. Theoretically, we rigorously establish a generalization bound and reveal how the predicted uncertainty, feature and network complexity affect this bound, and why setting $\boldsymbol{\alpha} = \mathbf{e} + \mathbf{1}$ can minimize it. Extensive experiments on standard visual and medical datasets demonstrate that VI-EDL achieves state-of-the-art performance, showing excellent performance in out-of-distribution detection, noise detection and autonomous driving scenario. The code is available in this https URL.
Machine Learning (cs.LG)
arXiv:2605.26477 [cs.LG]
(or arXiv:2605.26477v3 [cs.LG] for this version)
Text integral sub licența deschisă a publicației · 197 cuvinte
Subiecte conexe
Density-Informed Pseudo-Counts for Calibrated Evidential Deep Learning
arXiv · 20 iul. 2026
Meta-RL with Bayesian Linear Task Models
arXiv · 11 sept. 2026
Fisher Information based Stochastic Gradient Ascent for Online Learning of Dirichlet Process Mixture and Theory
arXiv · 28 iul. 2026
Generate images and video with vLLM-Omni on SageMaker AI – Part 2
AWS · 28 sept. 2026