Besides the examples provided in the user guides for the usage of models and quantifiers, the following Jupyter notebooks explain specific tasks:

  • Creating a Stochastic Model using the Sequential APIThis shows the simplest, and recommended, way to create an uncertainty aware DNN which is capable of calculating uncertainties and confidences based on point prediction approaches as well as on stochastic samples based approaches (e.g. MC-Dropout)

    Open In Colab View on Github

  • Convert a traditional keras Model into an uncertainty-aware modelThis shows how you can use any keras model you may have, which was not created through uncertainty wizard, into an uncertainty-aware DNN.

    Open In Colab View on Github

  • Create a lazily loaded and highly parallelizable Deep EnsembleThis show the fastest way to create an even faster implementation of the powerful Deep Ensembles - in a way which respects the fact that your PC and your GPU are powerful and your time is costly.

    Open In Colab View on Github

  • Multi-Device EnsembleThis shows an example on how to use uncertainty wizard lazy ensembles using multiple gpus in parallel.

    View on Github

    Note: As this example is only applicable to machines with two gpus, we do not provide a colab link or jupyter notebook, but instead a classical python script to be run locally.

More examples will be added when we get feedback from our first users about the steps they found non-obvious. In the meantime, you may want to check out the Complete API Documentation.