Machine Learning for Computational Economics

Instructor: Dejanir Silva (Purdue University) 🌐
Institution: EDHEC Business School

This course combines classical numerical methods in economics and finance with modern machine-learning approaches for solving and estimating dynamic models.

Students learn to connect discrete- and continuous-time dynamic programming with neural-network approximations and gradient-based optimization.

The course culminates with the Deep Policy Iteration (DPI) algorithm. Each module blends theory, implementation, and coding in Julia/Pluto.

Quick links

• 📝 Syllabus
• 📚 Lecture Notes (PDF)
• 🎓 Course Slides
• 💻 Interactive Notebooks
• 🧪 Replication Codes

Syllabus

For more details on the 2026 version of the course, you can view the syllabus here.

Lecture Notes

The main reference is the lecture notes developed for this course.
The notes go through the theory and implementation of the methods in detail and include many examples and guided implementations of the different methods in Julia.

You can view the lecture notes in PDF format here.

Course Slides

The course is divided into the following modules:

• Module 1 – Introduction
• Module 2 – Discrete-Time Methods
• Module 3 – Continuous-Time Methods
• Module 4 – Fundamentals of Machine Learning
• Module 5 – The Deep Policy Iteration Method

Interactive Notebooks

For each module, there is an interactive Pluto notebook that illustrates the main concepts and methods. Here are static versions of the notebooks for the different modules:

• Module 2 – Three Challenges
• Module 3 – Black-Scholes
• Module 4 – Fitting DNN
• Module 5 – Two Trees

Replication Codes

All replication codes for the course are available in the GitHub repository.
The code is organized by module in the src/ folder.