Notes

• Decentralized ML training: Multi-party collaboration, data privacy protection
• Traditional data science has structural biases, ignoring vulnerable groups' data
• Applications: Health analysis, edge group modeling, privacy recommendation, biometric recognition
• Model merging technology can analyze cross-identity features
• Technical advantages: Verifiable training, optimized edge computing efficiency
• Social value: Discovering overlooked social patterns through collaboration

• 去中心化ML训练：多方协作建模，保护数据隐私
• 传统数据科学存在结构性偏见，忽视弱势群体数据
• 应用：健康分析、边缘群体模型、隐私推荐、生物识别
• 模型合并技术可分析交叉身份特征
• 技术优势：可验证训练，优化边缘计算效率
• 社会价值：通过协作挖掘被忽视的社会模式