Deep Learning Analysis Updates and Reproducibility

[jjacobson95]

Numerous changes were made to allow the code to be more easily run. No changes were made to the actual model generation, just data formatting / robustness and others listed below.

Deep Learning & Transfer Learning Code:

• Code now works for updated data.
• Code now runs the same on all datasets.
  • There is no dataset specific code anymore - previously this existed due to decimals/ int/ string differences between datasets.
• Removed unused and commented out parts of the code.
• Made more robust to failures and now automatically checks for and creates output directories & files.
• Added doc strings and cleaned up the flow of the code

shared_input files:

• Added all of the required missing files (retrieved from James Moons deception home directory)

3_x_3 analysis scripts:

• Added / updated these so this analysis will be more straightforward in the future.

Extra files:

• Updated the calculate_statistics.py code so it can be run on a directory to retrieve and print all results at once
• Added remap_mpnst_ids.py file to replicate the method that James Moon modified the MPNST experiments file to allow it to work with the modeling code.
  • This remaps improve_sample_ids based on model type so the DL model can appropriately train/test the data.