了解和减轻图形对比学习的维度崩:一种非最大移除方法
Jiawei Sun1, Ruoxin Chen1, Jie Li1
1Department of Computer Science and Engineering, Shanghai Jiao Tong University, Shanghai, China.
图形对比学习 (GCL) 与维度崩作斗争. 我们的非最大移除GCL (nmrGCL) 方法理论上识别和减轻了这个问题,改善了图表表示学习性能.
科学领域:
- 人工智能的人工智能
- 机器学习 机器学习
- 图形表示学习学习学习图形表示学习
背景情况:
- 图形对比学习 (GCL) 优于无监督图形表示学习 (GRL),因为它最大限度地提高了图形视图之间的相互信息.
- GCL的一个关键限制是维度崩,其中嵌入被限制在一个低维子空间中,减少它们的表达力.
研究的目的:
- 从理论上分析GCL维度崩的原因.
- 提出一种新的方法,非最大去除GCL (nmrGCL),以解决维度崩.
主要方法:
- 理论分析确定图形聚合和图形卷积规则化是维崩的原因.
- 开发nmrGCL,在对比学习借口任务中删除正对中的突出维度.
主要成果:
- 拟议的nmrGCL方法有效地减轻了GCL的维度崩问题.
- 实验结果表明,nmrGCL在多个基准数据集上表现优于现有的最先进方法.
结论:
- GCL的维度崩可以归因于图形聚合和隐式规范化.
- nmrGCL提供了一个有前途的解决方案,以提高GCL在无监督图形表示学习中的表达力和性能.
更多相关视频
07:34Perceptual and Category Processing of the Uncanny Valley Hypothesis' Dimension of Human Likeness: Some Methodological Issues
Published on: June 3, 2013
07:12Development of a Gaze-Contingent Display Framework Designed for Perceptual and Oculomotor Research with Simulated Central Vision Loss
Published on: April 11, 2025
相关概念视频
Reducing Line Loss
With a step-up transformer at the source, the voltage is increased, thereby reducing the current in the transmission lines since power loss...
Survival Tree
Building a Survival Tree
Constructing a...
Difference from Background: Limit of Detection
The LOD indicates the presence or absence...
Quantifying and Rejecting Outliers: The Grubbs Test
Woodward–Hoffmann Selection Rules and Microscopic Reversibility
Routh-Hurwitz Criterion II
The first scenario occurs when a singular zero appears in the first column of the Routh table. This situation creates a division by zero issues. To resolve this, a small positive or negative number, denoted as epsilon (∈), is substituted for the zero. The stability analysis proceeds by assuming a sign for ∈. If ∈ is positive, any sign change in the first...
