Udemyintermediatefr

Apprentissage par renforcement avec Python - Partie 2

Learn AI & Reinforcement Learning with Python! Master TensorFlow & Keras for building artificial neural networks. Get hands-on experience with RL.

4.9(9)
168students
9h 57m
Certificate
Alexandre Bourrieau
Alexandre BourrieauEnseignant

Course Content

Apprentissage par renforcement avec Python - Partie 2 is an intermediate-level Udemy course taught by Alexandre Bourrieau. It runs 9h 57m, is taught in French, and includes a certificate of completion. Learners rate it 4.9 out of 5 from 9 reviews, with 168 students enrolled. It is a paid course, first published November 2022.

Master Reinforcement Learning: From Theory to Implementation with Python

Unlock the power of artificial intelligence with this advanced reinforcement learning course. Learn how to harness machine creativity by enabling AI to explore countless scenarios simultaneously – a capability far exceeding human limitations. This course builds upon a foundational understanding of reinforcement learning (Part 1), expanding your knowledge from finite to infinite environments, and from episodic to continuous tasks.

What You'll Learn:

  • Deep Reinforcement Learning: Dive into cutting-edge techniques using artificial neural networks with Keras and TensorFlow.
  • On-Policy & Off-Policy Algorithms: Master both online (on-policy) and offline (off-policy) reinforcement learning strategies.
  • Deep Q-Learning: Implement the powerful Deep Q-Learning algorithm for optimal decision-making in offline learning scenarios.
  • Deep-Sarsa: Explore Deep-Sarsa, a potent online learning method leveraging neural networks for strategy optimization.
  • Function Approximation: Learn to approximate both linear and non-linear functions using techniques like stochastic gradient descent and Fourier bases.
  • Continuous Control: Develop strategies for controlling agents in continuous environments.

Course Modules:

  1. On-Policy Prediction with Approximations: Introduction to linear function approximation for evaluating state values in continuous environments.
  2. Linear Function Approximation Construction: Advanced techniques for building linear function approximations using polynomial and Fourier bases.
  3. Non-Linear Function Approximation with Neural Networks: Utilize Keras and TensorFlow to build and apply artificial neural networks for approximating complex, non-linear functions.
  4. On-Policy Control with Approximations: Implement algorithms for obtaining optimal policies based on linear function approximations – the foundation for reinforcement learning control.
  5. Deep-Sarsa On-Policy Control: Optimize strategies using neural network-based function approximations in an online learning setting.
  6. Off-Policy Methods with Approximations: Address the challenges of stable and convergent strategy determination in offline learning with linear function approximations.
  7. Off-Policy Deep Q-Learning Control: Leverage Deep Q-Learning to optimize strategies in offline learning environments.

Who Should Take This Course?

  • Individuals with a basic understanding of Python programming.
  • Those with a background in Mathematics, particularly probability, will benefit from the theoretical explanations.
  • Familiarity with vector spaces is a plus.

Technical Requirements:

  • Python proficiency is recommended.
  • Access to a JupyterLab environment (e.g., Google Colab - free with a Gmail account).
  • All necessary resources are provided within the course.

Course Duration: Over 9 hours of video content.

Key Skills You'll Gain:

  • Implement reinforcement learning algorithms in Python.
  • Build and train artificial neural networks with Keras and TensorFlow.
  • Apply deep reinforcement learning techniques to solve complex problems.
  • Understand the differences between on-policy and off-policy learning.
  • Develop strategies for both online and offline reinforcement learning.
$ whoami — the devs behind tutorialsearch.io

Want AI in your product? Hire the team behind TutorialSearch.io

LLM features, RAG, agents and automation wired into a real product — built by the devs who shipped this site.