Deep Q learning Agent With Neural Network From Scratch

Author: Ariel Guerrero

Description

Learning Reinforced Learning, Neural Networks, Deep Learning, and many more exciting topics related to AI.

---

Single Agent

• learning about discrete and continuous environments that agents can act on randomly selecting appropriate actions to take in a given environment.

• code example: single_agent.ipynb

Q-Learning Agent

• learning about the concept of Q-Learning, which is a form of reinforcement learning that uses a Q-table to store the Q-values of the actions that the agent can take in a given environment.

• Q-Learning Algorithm

  • Q(s_t, a_t) = R_t+1 + gamma * max(Q(s_t+1))

  • s = state

  • a = action

  • R_t+1 = reward

  • gamma = discount factor

  • max(Q(s_t+1) = max Q value for all possible actions in state s_t+1

• reference: Q-Learning algorithm

• code example: q_agent.ipynb

Q-Learning Agent with Neural Network

• Using Q-Learning algorithm and a Sequential Model with two dense layers to dive deeper in learning about the concept of deep learning.

  

• code example: q_agent_nn.ipynb

Default Enviroment

• Frozen Lake