Introduction to TF Agents and Deep Q Learning (Reinforcement learning with TensorFlow Agents)

Wei Wei, a Developer Advocate for TensorFlow, introduces TF Agents and walks through how to use the Deep Q Learning model to solve the cartpole environment. Resources: TensorFlow Agents homepage → https://goo.gle/34i7MAI Train a Deep Q Network with TF Agents Tutorial → https://goo.gle/3oz26ZQ TF-Agent DQN example → https://goo.gle/3HxmXnM Reinforcement Learning Lecture Series 2021 (DeepMind x UCL) → https://goo.gle/3B6td3x Human Level Control Through Deep Reinforcement Learning (DQN) → https://goo.gle/3HE8PsO DeepMind Reverb: a framework for experience replay → https://goo.gle/3JgdMbF Opening up a physics simulator for robotics → https://goo.gle/34vBnq6 Chapters: 00:00 Introduction 00:23 What is TF Agents 1:38 TF Agents system overview 2:56 Deep Q Network (DQN) 4:10 Environment/Task 5:12 Define Q network 5:40 Define the DQN agent 5:49 Define the collect and eval policies 7:13 Set up the Reverb replay buffer 7:38 Define the replay buffer observer 7:54 Create the driver to collect experience 8:09 Inspect the experience trajectory 8:34 Run the training loop 8:59 Summary and references Watch more Reinforcement learning with TensorFlow Agents episodes → https://goo.gle/reinforcement-learning Subscribe to TensorFlow → https://goo.gle/TensorFlow Ask your questions on the TF Forum → https://goo.gle/discuss_tensorflow #TensorFlow #MachineLearning #ML product: TensorFlow - General; fullname: Wei Wei;