Written by Ali Payani
Based on paper : https://arxiv.org/pdf/1906.03523.pdf
The software is tested using python 3.6 and Tensorflow 1.13 Each experiment is a single Python file located in the root folder and all the experiments use the common library for dNL-ILP located in Lib folder.
List of current experiments:
-
Classification tasks: the only flag needed to be specified in the fold number in the crossvalidation scheme (0-4) for everything except for Mutagenesis which is 0-9
example: >python3 filename TEST_SET_INDEX=0 where filename can be any of {classify_mutagenesis.py, classifiy_cora.py, classifiy_uwcse.py, classifiy_imdb.py, classify_movielens.py}
-
other experiments can be envoked simply by running the corresponding py file
List of packages required for running the program: numpy, pandas, sklearn, tensorflow