Hubert Banville

Very nice example @gemeinl! I believe we had discussed using objects for windowers during the sprint? I think making full use of skorch would make a lot of sense.

We could instantiate the objects in session-level fixtures, and have function-level fixtures that do the deepcopy like here: https://stackoverflow.com/questions/53193932/how-to-avoid-mutating-objects-from-session-scope-fixtures-in-pytest

I like the idea. Only reason I see this could be problematic is if the concatenation of pd.Series is slow. I guess this is not likely to happen though as...

So to summarize, we have copies of `description` in `BaseDataset`, `WindowsDataset` and `BaseConcatDataset`. You're suggesting we don't actually copy `description` when creating `WindowsDatasets` or `BaseConcatDatasets`, but instead to provide getters...

Good point, people could just download a zip archive from https://github.com/peplin/pygatt and then use `pip install .` from inside the pygatt directory.

Thanks @anthonyromyn for testing out the repo and for reporting this! I've updated the code so that it is compatible with the latest version of braindecode, please let me know...

Hi @caowencai, thanks for submitting an issue! This is a good point. We'll add the relevant citations in Table 6 of the review. As for the links themselves, it might...