否定等号相似性:一种以理论为导向的方法,用于有效的表示学习
Takumi Nakagawa1, Yutaro Sanada2, Hiroki Waida3
1Department of Mathematical and Computing Science, Tokyo Institute of Technology, 2-12-1 Ookayama, Meguro-ku, Tokyo, 152-8550, Japan; RIKEN AIP, Nihonbashi 1-chome Mitsui Building, 15th floor, 1-4-1 Nihonbashi, Chuo-ku, Tokyo, 103-0027, Japan.
这项研究引入了一种新的无效化代数相似性 (dCS) 损失,以从杂的数据集中创建强大的机器学习表示. 在视觉和语音任务中,dCS损失提高了表示质量,超过了现有的方法.
科学领域:
- 机器学习 机器学习
- 数据科学数据科学数据科学
- 计算机视觉 计算机视觉
- 语音处理 语音处理
背景情况:
- 代表性学习对于各种任务的高效机器学习至关重要.
- 现实世界的数据集往往含有噪音,降低了学习到的表示的质量.
- 现有的方法在表达式学习过程中不足以解决噪声污染.
研究的目的:
- 开发一种学习强大的表示方法,可以抵抗数据集噪声.
- 提出一种新的损失函数,将denoising能力集成到表示学习中.
主要方法:
- 引入无效化代数相似性 (dCS) 损失,这是代数相似性损失的修改.
- 理论和经验验证dCS损失的消极性质.
- 为dCS损失开发可实施的估计器,并提供统计上的保证.
主要成果:
- 拟议的dCS损失表明与基线目标函数相比,性能优越.
- 经验证据证实了dCS损失在视觉和语音领域的有效性.
- 在训练数据中,学习的表示表现出对噪声的增强强性.
结论:
- 该dCS损失提供了一个有效的解决方案,用于从杂的数据集学习强大的表示.
- 这种方法在噪音条件下显著提高了表示质量和下游任务性能.
- 对于处理不完美的数据的实际机器学习应用程序,dCS损失提供了一个有价值的工具.
更多相关视频
09:09Foreign Accent and Forensic Speaker Identification in Voice Lineups: The Influence of Acoustic Features Based on Prosody
Published on: September 27, 2024
09:47Author Spotlight: Advancing Alzheimer's Research – Exploring Early Detection and Multi-Omics Approaches
Published on: December 15, 2023
相关概念视频
Linear Approximation in Frequency Domain
In contrast, nonlinear systems do not inherently possess these properties. However, for small deviations around an operating point, a nonlinear system can often be approximated as linear....
2D NMR: Homonuclear Correlation Spectroscopy (COSY)
Deconvolution
Deconvolution involves several mathematical techniques to derive the impulse response. One common approach is polynomial division. In this method, the input and output sequences are treated as coefficients of...
Direction Cosines of a Vector
2D NMR: Overview of Homonuclear Correlation Techniques
COSY90 is the standard two-dimensional (2D) COSY experiment that...
Dot Product
In engineering, the dot product of any two vectors is the product of the magnitudes of the vectors and the cosine of the angle between them. It is denoted by a dot symbol between the two vectors.
Consider a vehicle pulling an object along the ground using a rope. If the rope makes an angle with the horizontal axis, the work done can be calculated using the dot product of the force applied and the object's displacement.
The dot...
