thomas-jupyter-widget

Widget to display and interact with Bayesian Networks from thomas-core in JupyterLab.

This version only supports JupyterLab >= 3.

Table of contents

1. Regular Installation
2. Usage
3. Development installation

Regular Installation

The package can be installed through pip:

pip install thomas-jupyter-widget

Usage

In JupyterLab, try the following:

import thomas.core
from thomas.core import examples
from thomas.jupyter import BayesianNetworkWidget

# Load the Student network
Gs = examples.get_student_network()

# Display the widget
view = BayesianNetworkWidget(Gs, height=300)
display(view)

# Set evidence on the BN. This should update the probabilities throughout the
# network. Note that double-clicking on a state in the widget has the same
# effect.
Gs.set_evidence_hard('I', 'i1')

Development Installation

First ensure that nodejs (>=12) and yarn are installed. In debian, you can use the following if you have sudo privileges:

curl -sL https://deb.nodesource.com/setup_12.x | sudo bash -

# Running apt-get update is not necessary, since it's alread called in the
# above setup script.
sudo apt-get install -y nodejs

# Install yarn properly. The version of yarn that's available by default is
# just weird.
sudo apt-get remove cmdtest
sudo npm install --global yarn

Second, depending on your situation, you may want to create a virtual environment. For example, this might be useful if the command jupyter labextension develop --overwrite . fails because it cannot write to the system-wide extensions folder (e.g. /usr/local). Under debian/ubuntu, a virtual environment can be created as follows:

# Install the virtualenv package to make life easier
pip install virtualenv

# Create a virtual environment in the subdirectory '.venv'
virtualenv .venv

# Activate the virtual environment.
source .venv/bin/activate

After the preliminaries, install the package as follows:

# Clone the repository and move into the directory
git clone https://github.com/mellesies/thomas-jupyter-widget.git
cd thomas-jupyter-widget

# Checkout branch jupyter3
git checkout jupyter3

# Install the package in editable mode.
pip install -e .

# We need the latest version of JupyterLab, otherwise this will fail
pip install git+https://github.com/jupyterlab/jupyterlab

# link your development version of the extension with JupyterLab
jupyter labextension develop --overwrite .

# watch the source directory, automatically rebuilding when needed
yarn --cwd ./js run watch

Then, start JupyterLab like you normally would in a separate terminal (and don't forget to activate the virtual environment if you're using one ;-).