sanchit-agarwal / neuro-symbolic

Experiment code for dissertation report "Improving Classification using Neuro-Symbolic Algorithms"

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

neuro-symbolic

This repository contains the experiment code used in my dissertation submission "Improving classification using Neuro-Symbolic Algorithms"

Code includes six Jupyter notebooks:

  1. CNN.ipynb: Code for training and evaluating vanilla CNN model
  2. CNN_Noisy.ipynb : Code for training and evaluating vanilla CNN model on noisy data
  3. DeepProbLog B1.ipynb: Code for training and evaluating DeepProbLog model with B1 background knowledge
  4. DeepProbLog B2.ipynb: Code for training and evaluating DeepProbLog model with B2 background knowledge
  5. DeepProbLog B3.ipynb: Code for training and evaluating DeepProbLog model with B3 background knowledge
  6. DeepProbLog B4.ipynb: Code for training and evaluating DeepProbLog model with B4 background knowledge
  7. 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:

  1. torch
  2. torchvision
  3. matplotlib
  4. sklearn
  5. numpy
  6. pandas
  7. deepproblog

About

Experiment code for dissertation report "Improving Classification using Neuro-Symbolic Algorithms"


Languages

Language:Jupyter Notebook 97.4%Language:Prolog 2.6%