RLFinalProject

The repository contains code for experiements on using RL to find counterfactuals (or adversaries) for TABULAR DATA.

The code has been built upon from the skeletal code provided for assingments and my solutions for the same

1. env_setup.py : Built upon from enviroment file from Programming Assignment Ch 4,5,6,7
2. sarsa_lambda* : Built upon from skeletal code and own code written for Programming Assignment Ch12,13
3. Reinforce with NN: Adopted from skeletal code and own code written for Programming Assignment Ch12,13

To run the code, clone the repository and run different files to get different graphs as shown in the report. All files should be run with RLFinalProject as the home folder

1. Toy Example - Linear Case
   SARSA(lambda) with Tile Coding: run file sarsa_lambda_linear.py.
   Neural Net with Reinforce: run file nn_reinforce_linear.py
   

2. Toy Example - Non Linear Case
   SARSA(lambda) with Tile Coding: run file sarsa_lambda_non_linear.py.
   Neural Net with Reinforce: run file nn_reinforce_non_linear.py
   

3. Increased Dimensions: Linear Toy Example
   run file test_reinforce_increased_dimension.py
   

4. Pima Diabetes Dataset
   run file nn_reinforce_diabetes.py