SARFA

Link to the paper - SARFA paper

Working example - Example

Dependencies

1. Python3
2. Packages listed in requirements.txt

Using SARFA to interpret your RL Models

1. Clone Repository

git clone https://github.com/nikaashpuri/sarfa-saliency.git

2. Navigage to cloned repository

cd sarfa-saliency

3. Install minimal dependencies using pip (if you want to use SARFA with your Reinforcement Learning Models)

pip install -r requirements-minimal.txt

3. Install all dependencies using pip (if you want to use the examples included in the repository)

pip install -r requirements.txt

4. Use SARFA in your code

import sarfa_saliency
.....
.....

dict_q_vals_before_perturbation, original_action = RL_agent.q_values(state) # this is a dictionary, where the key is an action, and the value is the q-value of that action for this state 

# now perturb the state, this step depends on your domain 
# if you are using images or chess boards, you can use the perturbtations provided in our examples 
perturbed_state = perturb(state)

dict_q_vals_after_perturbation = RL_agent.q_values(perturbed_state)

saliency = sarfa_saliency.computeSaliencyUsingSarfa(original_action, dict_q_vals_before_perturbation, dict_q_vals_after_perturbation)

# that's all folks