PyTorch 1.x Reinforcement Learning Cookbook

How to do it...

We now develop double Q-learning to solve the Taxi environment as follows:

1. Import the required libraries and create an instance of the Taxi environment:

>>> import torch>>> import gym>>> env = gym.make('Taxi-v2')

2. Then, start defining the epsilon-greedy policy. We will reuse the gen_epsilon_greedy_policy function defined in the Developing Q-Learning algorithm recipe.

3. We then specify the number of episodes and initialize two variables used to track the length and total reward for each episode:

>>> n_episode = 3000>>> length_episode = [0] * n_episode>>> total_reward_episode = [0] * n_episode

Here, we simulate 3,000 episodes as double Q-learning takes more episodes to converge.

4. Next, we define the function that performs ...