Tutorial on Differentiable Probabilistic Programming for Agent-Based Models [ICAIF 2024]
Welcome to the tutorial on differentiable agent-based models!
Description
This tutorial will introduce a new paradigm for agent-based models (ABMs) that leverages automatic differentiation (AD) to efficiently compute simulator gradients. In particular, this tutorial will provide:
- An overview of vanilla AD and AD methods to differentiate through discrete stochastic programs.
- A walkthrough on how to build a differentiable economic ABM.
- State-of-the-art methods for calibrating differentiable ABMs.
Target audience
The tutorial is aimed at members of the ICAIF community who are interested in the deployment, calibration and analysis of large-scale agent-based models. No previous experience with automatic differentiation will be required to understand the tutorial material.
Outline and Schedule
Date: Friday 15th of November 2024
| Time | Session | Speaker |
|---|---|---|
| 08:00 - 08:40 | Automatic Differentiation for Agent-Based Models | Nicholas Bishop |
| 08:40 - 08:50 | Break | |
| 08:50 - 09:45 | Gradient-assisted calibration techniques for ABMs | Joel Dyer |
Materials
We have created a set of Jupyter notebooks which provide a practical walkthrough of the tutorial material. You can access the Google colab and nbviewer versions of each notebook via the links below.
- Automatic Differentiation: [Colab] [nbviewer]
- Differentiating Randomness: [Colab][nbviewer]
- Differentiable ABMs: [Colab][nbviewer]
- Variational Inference: [Colab][nbviewer]
Presentation slides, which supplement the notebooks above, are available here.
Presenters
Joel Dyer is a senior postdoctoral researcher at the University of Oxford’s Department of Computer Science and a Senior Research Fellow at the Oxford Institute for New Economic Thinking.
Nick Bishop is a postdoctoral researcher at the University of Oxford, working within the Department of Computer Science on problems at the intersection of machine learning and agent-based modelling.
Ani Calinescu is an Associate Professor at the University of Oxford’s Department of Computer Science, a Senior Research Fellow at the Oxford Institute for New Economic Thinking, and a Co-Investigator on a UKRI-funded project on Robust Large-Scale Agent-Based Modelling, and a Co-PI or Co-Investigator on several JPMC Faculty Research Awards.