MattStammers/q-FrozenLake-v1-8x8-noSlippery

Q-Learning Agent playing1 FrozenLake-v1

This is a trained model of a Q-Learning agent playing FrozenLake-v1 .

Usage

model = load_from_hub(repo_id="MattStammers/q-FrozenLake-v1-8x8-noSlippery-weak", filename="q-learning.pkl")

# Don't forget to check if you need to add additional attributes (is_slippery=False etc)
env = gym.make(model["env_id"])

To make this Q-learning agent work requires more extended training; otherwise the agent never successfully reaches the end goal and convergence does not take place.

In my case I found 50 million training steps sufficient with the following hyperparameters:

# Training parameters
n_training_episodes = 50000000  # Total training episodes
learning_rate = 0.99           # Learning rate

# Evaluation parameters
n_eval_episodes = 100        # Total number of test episodes

# Environment parameters
env_id = "FrozenLake-v1"     # Name of the environment
max_steps = 200             # Max steps per episode
gamma = 0.99                 # Discounting rate
epsilon = 0.1                # Ideal Episolon
eval_seed = []               # The evaluation seed of the environment

# Exploration parameters
max_epsilon = 1             # Exploration probability at start
min_epsilon = 0.05            # Minimum exploration probability
decay_rate = 0.0005            # Exponential decay rate for exploration prob