Local Interpretable Model Agnostic Explanations
PulseAugur coverage of Local Interpretable Model Agnostic Explanations — every cluster mentioning Local Interpretable Model Agnostic Explanations across labs, papers, and developer communities, ranked by signal.
2 天有情绪数据
-
新指标量化网络安全AI中的可解释性脆弱性
本文介绍了一种新颖的指标——可解释性脆弱性得分（Explanability Fragility Score），用于量化网络安全入侵检测系统中AI解释的不稳定性。研究表明，多重共线性（一种具有相关特征的统计问题）会显著放大解释方差，并导致特征重要性无法识别。为解决此问题，本文提出了两种缓解方法：CAA-Filtering和SHARP，旨在稳定AI解释，提高在安全关键应用中的可信度。
-
New methods enhance AI model explainability for images and tabular data
Researchers have developed two new methods for improving feature attribution in machine learning models. Spectral Integrated Gradients (SIG) uses singular value decomposition to create attribution paths that progress fr…
-
New framework enhances explainability for critical control systems
Researchers have developed a new framework called Hierarchical Causal Abduction (HCA) to make Model Predictive Control (MPC) systems more understandable. HCA combines physics-informed reasoning, optimization evidence fr…
-
Lime files for IPO, seeking $250M amid revenue growth and losses
Micromobility company Neutron Holdings, operating as Lime, has filed for an IPO with the SEC, aiming to raise up to $250 million. The company reported significant revenue growth, reaching $886.7 million in 2025, and has…
-
GRALIS framework unifies linear attribution methods for deep neural networks
Researchers have introduced GRALIS, a novel mathematical framework designed to unify various linear attribution methods used in Explainable AI (XAI). This framework establishes a canonical representation for attribution…
-
一种基于图增强知识蒸馏的双流视觉Transformer结合区域感知注意力用于胃肠道疾病分类及可解释AI
研究人员开发了一种新颖的双流深度学习框架，用于从医学影像中对胃肠道疾病进行分类。该系统采用教师-学生知识蒸馏方法，结合了用于全局上下文的Swin Transformer和用于细粒度特征的Vision Transformer。学生网络是一个紧凑的Tiny-ViT，在数据集1上达到了0.9978的高准确率，在数据集2上达到了0.9928，AUC为1.0000，同时还提供了更快的推理速度和更低的计算复杂度。可解释性分析证实了该模型依赖于临床…
-
New methods enhance low-light images using Retinex and Bayesian optimization
Researchers have developed FLARE-BO, an enhanced framework for improving low-light robotic vision. This new method expands upon a previous training-free approach by optimizing eight parameters, including gamma correctio…