The code here is implemented for the paper "Improving Certified Robustness via Statistical Learning with Logical Reasoning".
Each folder contains one corresponding experiment shown in the paper.
In specific:
-
The code for the experiment Road Sign Classification is contained in RoadSign.
-
The code for the experiment Information Extraction on Stock News is contained in NLP_MLN.
-
The code for the experiment Image Classification on PrimateNet Dataset is contained in PrimateNet.
-
The code for the experiment Image Classification on Word50 Dataset is contained in Word50-10.
-
The code for the experiment Image Classification on MNIST Dataset with Constructed Knowledge Rules is contained in MNIST.
We have detailed README files for guiding you how to run our code under each dataset subfolder.