readme and lfs

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+*.db filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,37 @@
+---
+library_name: pufferlib
+---
+# Play Pokemon with Reinforcement Learning
+
+## Project
+[Watch the Video](https://www.youtube.com/watch?v=DcYLT37ImBY)
+[GitHub](https://github.com/PWhiddy/PokemonRedExperiments)
+
+## Introduction
+The goal of this project is to use reinforcement learning to train an agent to play the game Pokemon Red. 
+The agent will learn to navigate through the game by interacting with the environment and receiving rewards or penalties based on its actions. 
+The ultimate objective is to have the agent achieve high scores in the game by mastering various tasks such as battling wild Pokémon, 
+catching them, and completing gym challenges.
+
+## Background
+Pokémon Red is a popular video game released in 1999 that has been widely enjoyed by people worldwide. 
+It was developed by Game Freak and published by Nintendo. 
+The game follows the story of a young trainer who sets out on a journey to become a Pokemon Master by capturing and training different creatures known as Pokemon. 
+The game features a unique battle system where players can command their Pokémon to perform various attacks and strategize against opponents.
+
+## Methodology
+To implement reinforcement learning for playing Pokemon, we will follow these steps:
+
+1. Define the problem and set up the environment: We need to define the objectives of the agent, which are to maximize the score and complete specific tasks in the game. 
+We also need to create a simulator that mimics the behavior of the original game, allowing us to test our algorithms without requiring access to the actual hardware.
+2. Choose a deep reinforcement learning algorithm: There are several RL algorithms available, but we will focus on PPO+LSTM due to its simplicity and effectiveness in solving complex problems. 
+PPO uses a neural network to approximate the action-value function, which maps states to expected future rewards.
+3. Train the agent: Once we have defined the problem and chosen an algorithm, we can start training the agent. In each episode, the agent selects an action based on its current state, takes it in the environment, 
+receives a reward or penalty, and updates its policy accordingly. This process continues until the desired level of performance is achieved.
+4. Test the agent: After training, we evaluate the agent's performance by testing it in the same environment used during training. We measure its success rate, average score, and other relevant metrics to determine how well it performed compared to human players.
+5. Refine the agent: Based on the results from testing, we may need to refine the agent's parameters, adjust the exploration strategy, or modify the reward function to improve its overall performance.
+
+## Expected Outcomes
+We expect that the trained agent will be able to successfully complete various tasks in Pokémon Red, such as defeating gym leaders, gaining levels and exploring the map. 
+By achieving these goals, the agent will demonstrate its ability to make decisions based on its experiences and adapt to new situations, making it a valuable tool for studying AI decision-making. 
+Additionally, the insights gained from this project could contribute to the development of more sophisticated RL models for tackling bigger game worlds.