This is an mini-course on "Deep Learning for Solving Dynamic Stochastic Models", held from Wednesday, May 22nd, 2024 2 - Friday, May 24th, 2024 at Central-German Doctoral Program Economics, University of Leipzig.
-
This mini-course is designed for Ph.D. students in economics and related disciplines. It introduces recent advancements in applied mathematics, machine learning, computational science, and the computational economics literature. The course focuses on solving and estimating dynamic stochastic economic models and performing parametric uncertainty quantification.
-
The lectures will concentrate on two machine learning methodologies: Deep Neural Networks and Gaussian Processes. These methods will be explored through applications in macroeconomics and climate-change economics.
-
The format of the lectures will be interactive and workshop-like, combining theoretical discussions with hands-on coding exercises. The coding will be conducted in Python and implemented on a cloud computing infrastructure.
Class enrollment on the Nuvolos Cloud
- All lecture materials (slides, codes, and further readings) will be distributed via the Nuvolos Cloud.
- To enroll in this class, please click on this enrollment key, and follow the steps.
- Nuvolos Support: support@nuvolos.cloud
- Basic econometrics.
- Basic programming in Python (see this link to QuantEcon for a thorough introduction).
- A brief Python refresher is provided under this link.
- A brief Python on Jupyter Notebooks is provided under this link
- Basic calculus and probability (The book Mathematics for Machine learning provides a good overview of skills participants are required to be fluent in).
Day 1, Wednesday, May 22nd, 2024
| Time | Main Topics |
|---|---|
| 09:30 - 11:00 | Introduction to Machine Learning and Deep Learning (part I) (2 x 45 min) |
| 11:00 - 11:30 | Coffee Break |
| 11:30 - 13:00 | Introduction to Machine Learning and Deep Learning (part II) (2 x 45 min) |
| 13:00 - 14:30 | Lunch Break |
| 14:30 - 16:00 | A hands-on session on Deep Learning, Tensorflow, and Tensorboard (2 x 45 min) |
Day 2, Thursday, May 23nd, 2024
| Time | Main Topics |
|---|---|
| 09:30 - 10:15 | Introduction to Deep Equilibrium Nets (DEQN) (2 x 45 min) |
| 10:15 - 11:00 | Hands-on: Solving a dynamic model with DEQNs (45 min) |
| 11:00 - 11:30 | Coffee Break |
| 11:30 - 12:15 | Hands-on: Solving a dynamic stochastic model with DEQNs (45 min) |
| 12:15 - 13:00 | Exercise: Solving a dynamic stochastic model by example (45 min) |
| 13:00 - 14:30 | Lunch Break |
| 14:30 - 15:15 | Introduction to a tuned DEQN library: solving a stochastic dynamic OLG model with an analytical solution (45 min) |
| 15:15 - 16:00 | Surrogate models part I: (for structural estimation and uncertainty quantification via deep surrogate models), with an example DSGE model solved with DEQN and pseudo-states (45 min) |
Day 3, Friday, May 24th, 2024
| Time | Main Topics |
|---|---|
| 09:00 - 10:00 | Surrogate models part II: (for structural estimation and uncertainty quantification via Gaussian process regression (60 min) |
| 10:00 - 10:30 | Creating GP-based surrogates from DSGE models (30 min) |
| 10:30 - 11:00 | Coffee Break |
| 11:30 - 12:15 | Introduction to the macroeconomics of climate change, and integrated assessment models (45 min) |
| 12:15 - 13:00 | Solving the (non-stationary) DICE model with Deep Equilibrium Nets (45 min) |
| 12:30 - 14:00 | Lunch Break |
| 14:00 - 15:30 | Putting things together: Deep Uncertainty Quantification for stochastic integrated assessment models; wrap-up of course (2 x 45 min) |
Lectures will be interactive, in a workshop-like style, using Python, scikit learn, Tensorflow, and Tensorflow probability on Nuvolos, a browser-based cloud infrastructure in which files, datasets, code, and applications work together, in order to directly implement and experiment with the introduced methods and algorithms.
- Simon Scheidegger (HEC, University of Lausanne)
Please cite Deep Equilibrium Nets, The Climate in Climate Economics, and Deep surrogates for finance: With an application to option pricing in your publications if this repository helps your research:
@article{https://doi.org/10.1111/iere.12575,
author = {Azinovic, Marlon and Gaegauf, Luca and Scheidegger, Simon},
title = {DEEP EQUILIBRIUM NETS},
journal = {International Economic Review},
volume = {63},
number = {4},
pages = {1471-1525},
doi = {https://doi.org/10.1111/iere.12575},
url = {https://onlinelibrary.wiley.com/doi/abs/10.1111/iere.12575},
eprint = {https://onlinelibrary.wiley.com/doi/pdf/10.1111/iere.12575},
year = {2022}
}
@article{10.1093/restud/rdae011,
author = {Folini, Doris and Friedl, Aleksandra and Kübler, Felix and Scheidegger, Simon},
title = "{The Climate in Climate Economics*}",
journal = {The Review of Economic Studies},
pages = {rdae011},
year = {2024},
month = {01},
issn = {0034-6527},
doi = {10.1093/restud/rdae011},
url = {https://doi.org/10.1093/restud/rdae011},
eprint = {https://academic.oup.com/restud/advance-article-pdf/doi/10.1093/restud/rdae011/56663801/rdae011.pdf},
}
@article{chen2023deep,
title={Deep surrogates for finance: With an application to option pricing},
author={Chen, Hui and Didisheim, Antoine and Scheidegger, Simon},
journal={Available at SSRN 3782722},
year={2023}
}
| Session # | Title | Screencast |
|---|---|---|
| 1 | First steps on Nuvolos | <iframe src="https://player.vimeo.com/video/513310246" width="640" height="400" frameborder="0" allow="autoplay; fullscreen; picture-in-picture" allowfullscreen></iframe> |
| 2 | Terminal intro | <iframe src="https://player.vimeo.com/video/516691661" width="640" height="400" frameborder="0" allow="autoplay; fullscreen; picture-in-picture" allowfullscreen></iframe> |