This repository contains examples using Gaussian Processes (GPs) to emulate agent-based models. It was developed as part of the project Uncertainty in agent-based models for smart city forecasts, funded by the Alan Turing Institute, in collaboration with the Data Assimilation for Agent-Based Modelling project.
Most of the files here use an agent-based model of bus movement as an example: the buses move between stops on a given route, picking up and dropping off passengers along the way.
The bus model simulation files are taken from the BusSim project, which is in turn part of the Data Assimilation for Agent-Based Models (DUST) project at the University of Leeds.
The original BusSim code is available from the DUST repository under an MIT licence; the relevant files are also included here.
For a quick look around the contents of the repository, click on the Binder badge link above. This will spin up a live instance of this repository in the cloud, with all packages installed for you. Changes made to the code there can be downloaded as separate files, but you cannot edit this repository directly from Binder - rather, it is a good option if you simply want to explore the notebooks.
For further development, create a fork of this repository, clone it and you will be able to edit and run the notebooks on your own computer.
Before running the code, you will need to install some Python packages. Generally, it is recommended to set up a new project in a virtual environment. To do so, run the following from the command line:
cd stochastic-gp
python3 -m venv venv
source venv/bin/activate
We will then use pip to install the required packages:
pip install --upgrade pip
pip install -r requirements.txt
pip install -r requirements-dev.txt
The requirements.txt file lists the packages needed to run the contents of the notebooks, while the requirements-dev.txt file lists utilities needed to use the project.
In this case, the contents of requirements-dev.txt relate to the Jupyter Lab installation.
To open Jupyter lab and explore the notebooks, run
jupyter lab
❓
Why is Jupyter Lab not included in the requirements.txt file?
This approach was taken to make it easy to set up Binder.
Binder detects a project's dependencies by searching for specific configuration files (here, requirements.txt and apt.txt).
Binder then builds a Docker image that contains all the project's required languages and dependencies.
As Binder itself provides the Jupyter Lab interface via Jupyter Hub, we do not need to install Jupyter Lab by default.
You can find out more about Binder here.
Note that some of the animated plots require a working installation of ffmpeg.
To install it, run
brew install ffmpeg # on Mac, with Homebrew
sudo apt install ffmpeg # on Linux, with apt