redefine default experiment settings
From my perspective, it is still hard to understand the default parameters for an experiment. To facilitate the process of understanding, it would be great to have a central collection of all defaults like it is already implemented for some parameters indicated as
DEFAULT_... in the
experiment_setup.py module. If this behaviour is applied to all customisable settings, the user could query for default arguments for an easy inspection. A arbitrary function
get_defaults() could return a dictionary with the parameter used as kwarg as name and its default value as dictionary's value. This would hopefully more intuitive than just reading the docs.