平衡混合型表格数据合成与扩散模型
Zeyu Yang1, Han Yu1, Peikun Guo1
1Department of Electrical and Computer Engineering Rice University.
这项研究引入了一种新的扩散模型,用于公平的合成表格数据生成,减轻训练数据集中的偏差,并与现有方法相比提高了10%以上的公平度量.
科学领域:
- 人工智能的人工智能
- 机器学习 机器学习
- 数据科学数据科学数据科学
背景情况:
- 扩散模型对于生成合成表格数据非常强大.
- 现有的模型经常继承并放大训练数据中存在的偏见.
- 有偏见的合成数据可能导致歧视性结果.
研究的目的:
- 开发一种新的表格扩散模型,生成公平的合成数据.
- 通过平衡目标标签和敏感属性的联合分布来减轻偏差.
- 确保高质量的合成数据生成,同时促进公平.
主要方法:
- 引入了一种新的表格扩散模型,其中包含了敏感指导.
- 平衡目标标签和敏感属性 (如性别,种族) 的联合分布.
- 经验评估使用公平度指标,如人口平价比率和均等赔率比率.
主要成果:
- 拟议的方法有效地减轻了培训数据中存在的偏差.
- 保持高质量的合成表格数据生成.
- 在公平性指标上表现优于现有方法,实现了超过10%的改进.
结论:
- 新的扩散模型成功生成了公平的合成表格数据.
- 该方法平衡了敏感属性和目标标签,减少了偏见.
- 这种方法为公平和高质量的合成数据生成提供了显著的进步.
更多相关视频
07:57An Experimental and Finite Element Protocol to Investigate the Transport of Neutral and Charged Solutes across Articular Cartilage
Published on: April 23, 2017
12:05A Simple, Robust, and High Throughput Single Molecule Flow Stretching Assay Implementation for Studying Transport of Molecules Along DNA
Published on: October 1, 2017
相关概念视频
Mechanistic Models: Compartment Models in Individual and Population Analysis
Model Approaches for Pharmacokinetic Data: Compartment Models
Two primary types of compartment models are recognized: mammillary and catenary. The more...
Physiological Pharmacokinetic Models: Blood Flow-Limited Versus Diffusion-Limited Models
Model Approaches for Pharmacokinetic Data: Distributed Parameter Models
The distributed parameter models are specifically designed to account for variations and differences in some drug classes. This model is particularly useful for assessing regional concentrations of anticancer or...
Compartment Models: Two-Compartment Model
Compartment Models: Single-Compartment Model
