Zhigang Feng · Professor of Economics
E-mail: z.feng2@gmail.com
This course covers how structural macroeconomic models are actually solved and analyzed on a computer. It runs across twelve topics: computation and programming foundations, including Python from scratch; numerical methods for dynamic models — dynamic programming, perturbation, projection, and accuracy assessment; high-performance and parallel computing; machine learning for macroeconomics; and heterogeneous-agent and life-cycle models. The organizing premise is that as AI takes over more of the implementation, the economist's edge moves to designing algorithms and validating results — the Research Architect.
Every lab runs on CPU alone — in GitHub Codespaces or on your own machine — so nothing is gated behind special hardware. In practice the machine-learning topics train noticeably faster on a GPU, so use one if you have access; none of them require it.
Prerequisites and where to go next. The course assumes intermediate macroeconomics — if you would like the theory firmer first, a previous offering of Advanced Macro Theory is recorded in full and free to watch. No programming background is needed: Topic 3, Programming Basics for Economists, teaches Python from scratch. The AI frontier that builds on this — reinforcement learning in depth, language models as economic measurement, and agentic research pipelines — is the companion course, AI for Economic Research: Dynamic Models, Language, and Agents. The two are designed as one sequence: classical methods → machine learning → agentic AI research.
This course has a complete video series: nine lectures, about nine and a half hours, recorded at the University of Miami in 2026 and free on Bilibili with no registration. Four earlier series — back to 2024 — cover the same ground at a slower pace.
Where to enter. New to computation: start at Topic 1 and do the setup lab. Already comfortable with Python and dynamic programming: Topics 7–9 are the classical core, and Topics 5–6 are the bridge to machine learning. Interested only in AI-assisted coding: Topic 3A stands alone. Heading for heterogeneous agents: Topics 11–12 are the destination.
| Stage | Focus | Topics |
|---|---|---|
| Foundations | AI for research; what quantitative macro needs; Python & PyTorch; AI-assisted coding | 1–3, 3A |
| Numerical core | Approximation, optimization, interpolation, discretizing shocks | 4 |
| Machine learning | Deep learning for economists; neural solutions to macro models | 5–6 |
| Classical methods | Dynamic programming, accuracy, perturbation, projection, HPC | 7–10 |
| Heterogeneity | Aiyagari and Krusell-Smith with ML; life-cycle OLG models | 11–12 |
This is the middle of three courses that run as one sequence. Before it, Advanced Macro Theory supplies the models — twenty-eight recorded lectures from the Solow model through recursive optimization, competitive equilibrium, asset pricing, and fiscal policy. After it, the AI frontier — large language models as economic measurement, retrieval-augmented generation, and running a research pipeline with agentic AI — is a course of its own, AI for Economic Research: Dynamic Models, Language, and Agents. This site is where the models get solved, and it is the subject of the textbook 机器学习与数量宏观经济学.
Macro Theory → Classical Methods → ML & Deep Learning → Agentic AI Research
You can enter at any stage. This site is the middle two.