Course Recordings · 课程录像

📺 The full course, on camera

This course was taught and recorded at the University of Miami in 2026nine lectures, about nine and a half hours, free to watch, no registration. The recordings follow the same twelve topics as this site, so a deck and a video exist for most of the material.

▶  Watch the 2026 series on Bilibili

2026 · University of Miami · 9 videos, ≈9 h 20 m

Watch in order, or jump to whichever topic you need. The right-hand column links the matching deck on the slides page.

#LectureLengthCovers
1 Introduction 1 h 10 m Topics 1, 2 — AI for economics research; what quantitative macro needs
2 Numerical Techniques for Quant Macro 1 h 11 m Topic 4 — approximation, optimization, root-finding
3 Numerical Techniques, continued 42 m Topic 4 — interpolation, quadrature, AR(1) discretization
4 Coding with AI for Quant Macro 32 m Topic 3A — the Research Architect workflow with Claude Code
5 Numerical Dynamic Programming 1 h 10 m Topic 7 — VFI and time iteration; closes by bridging to the Aiyagari model (Topic 11)
6 Advanced Numerical Methods & Accuracy Analysis 1 h 11 m Topic 8 — Howard, EGM, Euler equation errors
7 Perturbation and Projection Methods 1 h 15 m Topic 9 — Taylor expansions, Chebyshev, Smolyak
8 Machine Learning and Deep Learning for Macroeconomists 1 h 7 m Topics 5, 6 — PyTorch, then neural solutions to macro models
9 Solving DSGE Models via Deep Reinforcement Learning 1 h 5 m Topics 6, 11 — deep networks on Krusell-Smith, plus policy evaluation/improvement, which is sister-course RL material
Not every topic has a video. Topics 3 (Python & PyTorch), 10 (HPC) and 12 (OLG) were not recorded in the 2026 offering — their decks and, where available, labs stand on their own. For Topic 3, watch Lecture 3 of the 2025 series instead, which covers exactly that ground.

The "Covers" column above was checked against each lecture's own description rather than inferred from its title, so a few videos span more topics than their names suggest — Lecture 1 reaches into contraction mappings and Blackwell's conditions, and Lecture 2 opens on floating-point precision.

Earlier offerings

The same material taught at other institutions and in earlier years. Content overlaps with the 2026 series; the earlier series are longer and slower-paced, which some viewers prefer.

2025 · University of Miami — Quant Macro

8 videos — introduction, numerical methods, Python/PyTorch, three sessions on computing dynamic models, and two on heterogeneous agents.

The closest match to this syllabus after the 2026 series, and the only place two topics are taught on camera at length:

Lectures 4–6 ("Computation of Dynamic Models I–III") cover the ground of Topics 7–9.
https://space.bilibili.com/2142649036/lists/5800214

2025 · Iowa State University — AI/ML Workshop

13 videos — AI fundamentals, Python for research, PyTorch for deep learning, and applications.

The longest series, and the one that leans hardest on the AI side of the course.
https://space.bilibili.com/2142649036/lists/5800263

2025 · Summer intensive — AI/ML for Quant Macro

4 videos — a compressed four-session version of the course.

Start here if you want the shape of the whole thing in an afternoon.
https://space.bilibili.com/2142649036/lists/5800302

2024 · ML Macro

25 videos — the original long-form series, from floating point and optimization through machine learning for macro.

Broken into short focused videos, which makes it the easiest to dip into for one specific method.
https://space.bilibili.com/2142649036/lists/3144093

Advanced Macro Theory — 高级宏观经济理论

Separate from the computational series above, a previous offering of Advanced Macroeconomic Theory is recorded in full — the growth model, competitive equilibrium, overlapping generations, asset pricing, fiscal policy, and uncertainty. It is the theory this course puts on a computer, and it is taught in Chinese. See Advanced Macro Theory, or go straight to the series on Bilibili.

The textbook

The written counterpart to these lectures is 机器学习与数量宏观经济学 (Machine Learning and Quantitative Macroeconomics) by Zhigang Feng, which bridges classical numerical methods and deep learning for dynamic economic models.

Where this course sits

These recordings and this site are the foundation stage. The frontier — language models as measurement, retrieval, and agentic research pipelines — is the sister course, AI for Economic Research: Dynamic Models, Language, and Agents.

Classical Methods ML & Deep Learning Agentic AI Research

This site is the first two stages.