QAgentLib

A specialized reinforcement learning package inspired by and building upon the core concepts of Agentic Learn Pro, with targeted features for specific domains.

Overview

This library provides a clean, modular, and well-documented implementation of Q-learning with specialized extensions for targeted applications. It demonstrates the influence and advancement inspired by the original Agentic Learn Pro project, while focusing on specific use cases and advanced reinforcement learning techniques.

Attribution

This package is an independent implementation inspired by the foundational work of Agentic Learn Pro, created by Stephanie Ewelu. We acknowledge and appreciate the significant contributions of Agentic Learn Pro to the field of reinforcement learning and agentic AI.

Installation

pip install QAgentLib

Usage

Basic Q-Learning

from QAgentLib.agent import QLearningAgent
from QAgentLib.environment import SimpleEnv

env = SimpleEnv()
agent = QLearningAgent(state_space=env.state_space, action_space=env.action_space)

for episode in range(100):
    state = env.reset()
    done = False
    while not done:
        action = agent.choose_action(state)
        next_state, reward, done = env.step(action)
        agent.learn(state, action, reward, next_state)
        state = next_state
    agent.decay_exploration()
print("Q-table after training:", agent.q_table)

Targeted Q-Learning with Advanced Features

from QAgentLib.agent import TargetedQLearningAgent
from QAgentLib.environment import GridWorldEnv

# Create a grid world with obstacles
env = GridWorldEnv(
    width=5, 
    height=5, 
    start_pos=(0, 0), 
    goal_pos=(4, 4),
    obstacles=[(1, 1), (2, 1), (3, 1), (1, 3), (2, 3), (3, 3)]
)

# Create a targeted agent for grid navigation
agent = TargetedQLearningAgent(
    state_space=[(x, y) for x in range(env.width) for y in range(env.height)],
    action_space=[0, 1, 2, 3],  # up, right, down, left
    target_domain="grid_navigation"
)

# Training loop with experience replay
for episode in range(200):
    state = env.reset()
    total_reward = 0
    done = False
    
    while not done:
        action = agent.choose_action(state)
        next_state, reward, done, info = env.step(action)
        agent.learn(state, action, reward, next_state)
        
        # Replay experiences periodically
        if episode > 10 and episode % 5 == 0:
            agent.replay_experience(batch_size=10)
            
        state = next_state
        total_reward += reward
        
    # Adjust learning rate based on performance
    agent.adaptive_learning_rate()
    agent.decay_exploration(0.95)
    
    if episode % 20 == 0:
        metrics = agent.get_performance_metrics()
        print(f"Episode {episode}, Avg Reward: {metrics['avg_reward']:.2f}")
        print(env.render())

Multi-Objective Reinforcement Learning

from QAgentLib.agent import TargetedQLearningAgent
from QAgentLib.environment import MultiObjectiveEnv

# Create environment with 3 competing objectives
env = MultiObjectiveEnv(
    num_objectives=3,
    objective_weights=[0.5, 0.3, 0.2]  # Prioritize first objective
)

# Create agent for multi-objective optimization
agent = TargetedQLearningAgent(
    state_space=env.state_space,
    action_space=list(range(env.action_space)),
    target_domain="multi_objective"
)

# Training loop
for episode in range(100):
    state = env.reset()
    done = False
    
    while not done:
        action = agent.choose_action(state[0])  # state is (state_index, objective_values)
        next_state, reward, done, info = env.step(action)
        agent.learn(state[0], action, reward, next_state[0])
        state = next_state
        
    if episode % 10 == 0:
        metrics = agent.get_performance_metrics()
        print(f"Episode {episode}, Performance: {metrics['avg_reward']:.2f}")
        print(f"Objective values: {info['objective_values']}")

Development

For development, clone the repository and install in editable mode:

git clone https://github.com/your_username/QAgentLib.git
cd QAgentLib
pip install -e .

Testing

To run tests:

pytest

License

This project is licensed under the MIT License - see the LICENSE file for details.