Stanford AI 课程地图 EN GitHub ↗
← 返回课程地图
Stanford · 2026—27 MS&E 318 统计与优化 ★★★★★

核心课程 · 统计与优化

安全与受约束的 AI

Safe and Constrained AI

安全约束 AI:约束优化、safe RL、投影网络、LLM alignment。

3 学分 2026–27 未排课 Letter or Credit/No Credit ★★★★★
安全约束优化RL对齐

2026-27 学年未在 ExploreCourses 排课。

课程定位

MS&E 318 是一门博士级 seminar,讨论如何让 AI 系统不仅强大,而且“可证明地安全”。课程覆盖经典约束优化、safe RL(trust region、Lyapunov、reachability、shielding)、ML 与优化混合(projection networks、solver-in-the-loop)以及 LLM alignment,应用横跨机器人、金融、医疗、能源。

主要内容

  1. 约束优化:Lagrangian、robust/stochastic programming
  2. Safe RL:trust region、Lyapunov、reachability、shielding
  3. 混合方法:projection networks、solver-in-the-loop
  4. LLM 对齐:微调、模型编辑、工具使用、交互式对齐
  5. 理论结果:可行性、稳定性、保证
  6. 应用与项目

先修

优化水平达到 CME 307 或 EE 364A。

考核与工作量

论文讨论 + final project。具体以当季官网为准。

适合谁

安全 RL、控制、优化、alignment 方向的研究生;想做“有保证的 AI”的人。

学习建议

先补凸优化和控制理论。课程偏理论,项目要选一个能证明/度量安全性的问题。

相关课程

官方课程描述(英文原文)

How can we design AI systems that are not only powerful but also provably safe and trustworthy? This advanced PhD seminar surveys algorithmic methods to enforce hard constraints in machine learning, reinforcement learning, and generative AI. Topics include classical constrained optimization (Lagrangian methods, robust and stochastic programming), safe reinforcement learning (trust regions, Lyapunov functions, reachability, shielding), hybrid ML-optimization methods (projection networks, solver-in-the-loop architectures), and alignment strategies for large language models (fine-tuning, model editing, tool use, and interactive alignment). Each week highlights a key theoretical result alongside state-of-the-art research, with applications spanning robotics, finance, healthcare, energy, and language models. Students will critically assess the strengths and limitations of these methods and develop final projects that apply or extend them in real-world domains. Prerequisites: optimization at the level of CME 307 or EE 364a.