This page will soon be deactivated—explore our new, faster, mobile-friendly site, now centralized in MyWorkspace!

Connecting the world and beyond

  •  
PP 2026

ITU-T Recommendations

Search by number:
Others:
Skip Navigation Links
Content search
Advanced search
Provisional name
Equivalent number
Formal description
Study Groups tree viewExpand Study Groups tree view

ITU-T Q.4081 (01/2026)

عربي | 中文 | English | Español | Français | Русский
监测包括IMT-2020在内的未来网络中机器学习/人工智能的方法和指标
机器学习(ML)/人工智能(AI)监测是指,在生产环境中对ML/AI系统进行持续、实时地跟踪和观察。ML/AI监测旨在评估ML/AI模型的性能,以确定其是否处于有效运行状态。当ML/AI模型出现部分性能下降时,应采取适当的维护措施以恢复其性能。ML/AI模型是基于历史数据和对运行环境的假设进行训练的。但环境是动态的,这种动态变化可能引发数据漂移和概念漂移等现象,进而导致模型退化,即模型的预测准确性或决策质量随时间推移而下降。因此,应对ML/AI模型的环境和运行状态进行监测,以确定是否应对模型进行更新。为应对各类导致性能下降的问题,需定义并监测一组相应的参数和事件。因此,根据具体用例、数据特征和业务需求选择合适的监测策略,是确保ML/AI系统长期可靠性和有效性的关键步骤。ITU-T Q.4081建议书提供了关于ML/AI监测方法和指标的指南和参考。
Citation: https://handle.itu.int/11.1002/1000/16690
Series title: Q series: Switching and signalling, and associated measurements and tests
  Q.3900-Q.4099: Testing specifications
  Q.4060-Q.4099: Testing specifications for IMT-2020 and IoT
Approval date: 2026-01-13
Provisional name:Q.MMAI
Approval process:AAP
Status: In force
Maintenance responsibility: ITU-T Study Group 11
Further details: Patent statement(s)
Development history
[12 related work items in progress]