Jove
Visualize
联系我们
JoVE
x logofacebook logolinkedin logoyoutube logo
关于 JoVE
概览领导团队博客JoVE 帮助中心
作者
出版流程编辑委员会范围与政策同行评审常见问题投稿
图书馆员
用户评价订阅访问资源图书馆顾问委员会常见问题
研究
JoVE JournalMethods CollectionsJoVE Encyclopedia of Experiments存档
教育
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab Manual教师资源中心教师网站
使用条款与条件
隐私政策
政策

相关概念视频

您也可能阅读

相关文章

通过共同作者、期刊和引用图与本文相关的文章。

排序
Same author

<b>Erratum: KAI WANG, FU SHU, LING LI, WEN-JIE DONG, HAO-NAN MU, MING SU, WEN-BO ZHU, MENG-FEI ZHANG, KE-JI GUO & JING CHE (2025) A New Cryptic Species of <i>Calotes</i> (Reptilia: Squamata: Agamidae) from Southeastern Xizang Autonomous Region, China. <i>Zootaxa</i>, 5609 (2), 275-285.</b>

Zootaxa·2026
Same author

Youth educational mobility and the rural family in China.

Research in education (University of Manchester)·2026
Same author

PDGFRα<sup>+</sup>/Integrin α2<sup>+</sup> Fibroblasts Orchestrate Tumor Budding in Oral Squamous Cell Carcinoma via Mechano-Metabolic Symbiosis: E-Cadherin/Integrin α2β1 Adhesion and Mitochondrial Transfer.

Advanced science (Weinheim, Baden-Wurttemberg, Germany)·2026
Same author

Predicting range expansion of three tropical fruit flies and risks to pome and stone fruit production in temperate regions.

Journal of economic entomology·2026
Same author

Aligning Chemical Kinetics with Crystallization Enables Millimeter-Scale Single Crystals of Conductive MOFs.

Journal of the American Chemical Society·2026
Same author

Reframing Ultrasound-Based Dyspnea Models for Idiopathic Pulmonary Fibrosis.

Journal of ultrasound in medicine : official journal of the American Institute of Ultrasound in Medicine·2026

相关实验视频

Updated: May 13, 2025

Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness
03:14

Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness

Published on: December 6, 2024

462

用METER进行知识增强,参数高效的转移学习,用于医学视觉语言任务.

Xudong Liang1, Jiang Xie1, Jinzhu Wei2

  • 1School of Computer Engineering and Science, Shanghai University, Shanghai, China.

Journal of biomedical informatics
|May 10, 2025
PubMed
概括
此摘要是机器生成的。

使用METER (KPL-METER) 增强知识的参数效率转移学习通过整合外部知识和参数效率微调 (PEFT) 方法来改善医疗视觉语言任务. 与传统的PEFT方法相比,这种方法提高了性能并降低了计算成本.

关键词:
深度学习是一种深度学习.医学分析 医学分析具有参数效率的微调.视觉语言是一种视觉语言.

更多相关视频

Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
04:48

Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography

Published on: November 30, 2022

2.6K
Development of a Gaze-Contingent Display Framework Designed for Perceptual and Oculomotor Research with Simulated Central Vision Loss
07:12

Development of a Gaze-Contingent Display Framework Designed for Perceptual and Oculomotor Research with Simulated Central Vision Loss

Published on: April 11, 2025

206

相关实验视频

Last Updated: May 13, 2025

Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness
03:14

Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness

Published on: December 6, 2024

462
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
04:48

Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography

Published on: November 30, 2022

2.6K
Development of a Gaze-Contingent Display Framework Designed for Perceptual and Oculomotor Research with Simulated Central Vision Loss
07:12

Development of a Gaze-Contingent Display Framework Designed for Perceptual and Oculomotor Research with Simulated Central Vision Loss

Published on: April 11, 2025

206

科学领域:

  • 人工智能的人工智能
  • 机器学习 机器学习
  • 医疗信息学 医疗信息学

背景情况:

  • 为医疗任务进行预训练模型的完整微调是计算上昂贵的.
  • 参数效率微调 (PEFT) 方法降低成本,但由于领域差距,可能表现不佳.
  • 将通用视觉语言 (VL) 模型适应专业医疗领域仍然具有挑战性.

研究的目的:

  • 为医疗VL任务使用METER (KPL-METER) 提出知识增强的参数效率转移学习.
  • 通过将外部医学知识整合到PEFT中来提高模型性能.
  • 解决一般预训练模型和医疗应用之间的领域差距.

主要方法:

  • 开发了KPL-METER,将PEFT与外部领域特定知识相结合.
  • 引入共享适配器 (SAdapter) 用于多式联运分支机构,以保持单式联运特征和跨式联运一致性.
  • 使用统一医学语言系统 (UMLS) 实施了一种新的知识提取和无参数建模策略.
  • 添加了图像和文本编码器的适配器,以进一步完善单模特征.

主要成果:

  • 在使用更少的参数的情况下,KPL-METER在医疗VL任务上表现优于其他PEFT方法.
  • 采用医疗定制编码器的KPL-METER-MED,与以前的医疗领域模型相比,在调节参数较少的情况下实现了更高的性能.
  • 提出的方法有效地弥合了领域的差距,并提高了性能.

结论:

  • KPL-METER架构成功地将一般VL模型适用于医疗应用.
  • 知识提取和融合方法通过结合医疗领域特定的知识,显著提高了性能.
  • 该研究为医疗VL任务提供了高效和有效的解决方案.