Skip to content

Correcting a set of small (and less small) bugs

Cecile Le Sueur requested to merge corrections_code into main

!!!!! one major change!!!! I forgot a minus in the MLL definition of the Nextflow implementation!

The results of the paper were obtained with the previous code, which is without this mistake. With this correction in the Nextflow implementation, we now recover the same results as with the original implementation, i.e. same results than in the paper.

Note that in "predict_joint_HGPM" function (in predict_HGPM.py), the value_mll is correctly set to -mll, i.e the loss which is the negative of the marginal likelihood.

Additional minor changes:

  • change "minimise_required_memory_space" in user_params.config from "True" to "true".
  • check that the control condition is defined as one of the Level_2 values
  • add "optional" to nextflow outputs such that optional outputs (like scaled data or plots) do not break the pipeline if inexistant
  • correct the argument "PlotSave" in "parameters.txt" as it was not taken into account in the code.

Merge request reports