Welcome to the seminar!

You can use this website to access materials related to the workshop.

Schedule

Day Time Topic
Wed 9:00-10:30 Introduction, Deep Learning
Wed 11:00-12:30 Deep Learning, Generative Neural Networks
Wed 14:00-15:30 Practical exercise
Wed 16:00-17:30 Amortized Bayesian Inference with BayesFlow
Thu 9:00-10:30 BayesFlow demonstration
Thu 11:00-12:30 Practical exercise

Slides

Here you can access the presentation slides

  1. Introduction
  2. Deep Learning
  3. Generative Neural Networks
  4. Amortized Bayesian Inference with BayesFlow

Exercises

Use the Exercises tab in the navigation bar to access exercise notebooks.

Projects

Use the Projects tab in the navigation bar to access the project descriptions. See https://quantitative-thinking.eu/mobilities/seminar-2025/ for all projects.

Environment

The exercises and projects require Python 3.10 – 3.12, and installing necessary libraries. The recommended steps using conda:

Terminal
# create a "bayesflow-seminar" conda environment with Python 3.11
conda create --name bayesflow-seminar python=3.11.11

# activate the environment
conda activate bayesflow-seminar

# kernel for running the jupyter notebooks
conda install ipykernel --update-deps --force-reinstall

# install packages
pip install tensorflow
pip install git+https://github.com/bayesflow-org/bayesflow@main

When running the exercise notebooks, remember to use the correct environment!

Note: bayesflow can run with jax or pytorch instead of tensorflow. If you prefer to use those as a backend, you can install them as well. Remember to set the correct environment before loading keras, e.g., os.environ["KERAS_BACKEND"] = "jax". Note that the examples of generative neural networks do require tensorflow regardless.