相关实验视频

Updated: Jul 6, 2026

Basics of Multivariate Analysis in Neuroimaging Data

Basics of Multivariate Analysis in Neuroimaging Data

Published on: July 24, 2010

MuSIA:利用多源信息融合与异常激活进行分布外检测

Heng-Yang Lu¹, Xin Guo², Wenyu Jiang³

¹School of Artificial Intelligence and Computer Science, Jiangnan University, Wuxi, China; The Laboratory for Advanced Computing and Intelligence Engineering, Wuxi, China.

Neural networks : the official journal of the International Neural Network Society

|April 4, 2025

概括

这项研究引入了MuSIA,这是一种用于分布外 (OOD) 检测的新方法,它将多源信息与异常激活融合在一起. 通过减少OOD样本的错误阳性率,MuSIA显著提高了深度学习模型的稳定性.

科学领域:

人工智能的人工智能
机器学习机器学习
计算机视觉计算机视觉

背景情况:

在开放世界的场景中,OOD检测对于可靠的深度学习至关重要.
现有的OOD检测方法由于依赖单一来源信息和异常激活而表现不佳.
深度学习模型需要强大的机制来识别和处理不熟悉的数据.

研究的目的:

提出MuSIA (多源信息融合与异常激活) 进行增强的OD检测.
通过整合来自多个来源的信息和捕获异常激活来提高OOD检测的性能.
提高对OOD样本的深度学习模型的可靠性和稳定性.

主要方法:

MuSIA整合了来自多个来源的信息,以捕捉全面的特征.
该方法专门针对并利用OOD数据中的异常激活模式.
实验使用六个不同的OOD数据集和六个预训练模型 (例如ViT,RepVGG,Diet) 进行.

主要成果:

与最先进的 (SOTA) 方法相比,MuSIA显示出更高的性能.
提出的方法在95%的假阳性率 (FPR95) 中平均降低了7.78%的假阳性率.
废弃性研究证实了每个成分的有效性,特别是多源融合和异常激活捕获之间的协同作用.

结论:

MuSIA在OOD检测能力方面提供了显著的进步.
多源信息和异常激活的融合是提高OOD检测性能的关键.
拟议的方法提高了深度学习模型在现实应用中的稳定性和可靠性.

关键词:

计算机视觉计算机视觉计算机视觉图像的分类图像的分类.在分销之外的检测检测

更多相关视频

Using Informational Connectivity to Measure the Synchronous Emergence of fMRI Multi-voxel Information Across Time

Using Informational Connectivity to Measure the Synchronous Emergence of fMRI Multi-voxel Information Across Time

Published on: July 1, 2014

Concurrent EEG and Functional MRI Recording and Integration Analysis for Dynamic Cortical Activity Imaging

Concurrent EEG and Functional MRI Recording and Integration Analysis for Dynamic Cortical Activity Imaging

Published on: June 30, 2018

相关概念视频

¹H NMR: Interpreting Distorted and Overlapping Signals

¹H NMR: Interpreting Distorted and Overlapping Signals

Spin systems where the difference in chemical shifts of the coupled nuclei is greater than ten times J are called first-order spin systems. These nuclei are weakly coupled, and their chemical shifts and coupling constant can generally be estimated from the well-separated signals in the spectrum.
As Δν decreases and the signals move closer, the doublets appear increasingly distorted. The intensities of the inner lines increase at the cost of those of the outer lines as the signals are slanted or...