conv_qsar_fast
conv_qsar_fast copied to clipboard
Compatible with Py3?
This is a very interesting cheminformatrics approach. I was trying to learn and use the source code. I managed to change a couple of errors reported by Python 3 and made the code running with keras 2.1.6 and theano 1.0.3. But the example cases always stopped at 10th epochs with early stopping. It did not appear the training performed correctly and the testing performance was very poor compared to what was reported in the paper. Any suggestion to make it compatible with Py3? Thanks.
If anyone else is interested (it's already 2020) I pushed a (semi-)working code to https://github.com/Lamiane/conv_qsar_fast The results for TOX are similar as in the paper and it runs on python 3
I also made a working version of svm_cv.py and main_cv.py for python 3. Codes are available at https://github.com/LanceKnight/conv_qsar_fast. I tested this on the De-aq dataset.