This repository contains the experiment code used in my dissertation submission "Improving classification using Neuro-Symbolic Algorithms"
Code includes six Jupyter notebooks:
- CNN.ipynb: Code for training and evaluating vanilla CNN model
- CNN_Noisy.ipynb : Code for training and evaluating vanilla CNN model on noisy data
- DeepProbLog B1.ipynb: Code for training and evaluating DeepProbLog model with B1 background knowledge
- DeepProbLog B2.ipynb: Code for training and evaluating DeepProbLog model with B2 background knowledge
- DeepProbLog B3.ipynb: Code for training and evaluating DeepProbLog model with B3 background knowledge
- DeepProbLog B4.ipynb: Code for training and evaluating DeepProbLog model with B4 background knowledge
- Training plot.ipynb: Code for plotting training time recorded for all the models (excluding the noisy ones)
Following dependencies must be installed prior to running the experiemnts:
- torch
- torchvision
- matplotlib
- sklearn
- numpy
- pandas
- deepproblog