arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2104.13621cs.LGstat.ML
收藏

MLDemon:机器学习系统的部署监控

MLDemon: Deployment Monitoring for Machine Learning Systems

Stanford University(斯坦福大学);Harvard University(哈佛大学)

Antonio Ginart, Martin Zhang, James Zou

摘要:

ML系统的部署后监控对于确保可靠性至关重要,特别是当新用户输入可能与训练分布不同时。在这里,我们提出了一种新的方法,MLDemon,用于ML部署监控。MLDemon集成了未标记的数据和少量的按需标签,以生成ML模型在给定数据流上的当前性能的实时估计。根据预算限制,MLDemon决定何时获取额外的、可能成本高昂的专家监督标签来验证模型。在具有不同分布漂移和模型的时态数据集上,MLDemon优于现有方法。此外,我们提供的理论分析表明,MLDemon是极大极小率最优的广泛的一类分布漂移。

英文摘要:

Post-deployment monitoring of ML systems is critical for ensuring reliability, especially as new user inputs can differ from the training distribution. Here we propose a novel approach, MLDemon, for ML DEployment MONitoring. MLDemon integrates both unlabeled data and a small amount of on-demand labels to produce a real-time estimate of the ML model's current performance on a given data stream. Subject to budget constraints, MLDemon decides when to acquire additional, potentially costly, expert supervised labels to verify the model. On temporal datasets with diverse distribution drifts and models, MLDemon outperforms existing approaches. Moreover, we provide theoretical analysis to show that MLDemon is minimax rate optimal for a broad class of distribution drifts.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑