qSAR-net

---

Packages:

• python 3.5.2
• ipython 5.1.0
• notebook 4.2.3
• numpy 1.11.2
• openbabel 2.4.1 and/or rdkit 2016.03.4 (+ pillow=3.4.1)
• pandas 0.19.0
• scikit-learn 0.18
• scipy 0.18.1
• seaborn 0.7.1
• tensorflow 0.10.0rc0

 

for GPU-enabled version of TensorFlow:

• cudatoolkit 7.5
• cudnn 5.1
• tensorflow 0.11.0 (from mwojcikowski channel)

Files:

• saved_sets/history.txt - contains confusion matrices and max validation rates from ten runs of the best network on the training set; 'cm':confusion_matrices, 'vr':max_validation_rates; it is used to find classes with very high and very low mean True Negative Rates