GitLab

Update github sync workflow, switch to docker, add gitlab CI for building docker container and push to registry

other wise version mismatch leading to https://github.com/yigbt/deepFPlearn/pull/11#issuecomment-1113750598

Jana/example

This reverts commit 07b11394.

from requirements and environment

and add me as co-author in setup.py

Patrick

TestSize 0.0 needs to be handled in the FNNs which it currently is not.

Add bias to last layer in SNN

- Use BinaryAccuracy
- Change target type to short as it should be

- Use StratifiedKFold instead of KFold to also stratify each single fold
- Fix bug that would always down-sample when running uncompressed FNN
- Fix all log outputs and string formattings
- Add forgotten y_one_hot in a conditional branch
- Make unknown loss function fail explicitly
- Improve collecting the performance metrics for all targets and folds

God, I hope I didn't break something. Looks good so far.

That was the reason for having this double logging of entries in the console.
When running the main program, the logger is set up by a call to "createLogger". If you need it in the REPL, you can also call this function providing an output file.

one-hot encode y

Fix stratification of FNN samples
Fix stupid error in activation selection