Introduction and Hands-on workshop with Reinforcement Learning

Reinforcement Learning is a well-known Learning method that mimics exactly how the human learns. GYM Library is a standard API for reinforcement learning and it provides environments and interfaces to solve real-life problems and play around with Algorithms.

Transcript

1. Introduction and Hands-on workshop with Reinforcement Learning Presented by: Krupa Galiya Monday, 31 October, 2022
2. Agenda 1. Introduction of Reinforcement Learning a. How RL Differs with Supervised and Unsupervised Learning? b. Real-life Examples 2. Agents and Environment 3. Learn to Formulate a basic Reinforcement Learning problem 4. Open-source RL Libraries 5. Hands-on: Solve a problem with RL algorithm 6. Use-cases and RL Resources
3. How RL differs? Image Source
5. Image Source Environment: house Agent: Dog Action: Muscle Contractions Observations: Sight, smell Reward: Food
6. Image Source Environment: Surface Agent: Child Action: Trying to balance the body weight, Step a foot ahead Observations: Seeing other to walk, Which and how to step a foot ahead Reward: Claps
7. • Typically referred as an Agent interacting with an unknown environment. • Agent learns from its environment sequentially through actions and the feedback it gets. • Superpower: learns very complex behaviors without requiring labeled training data, and can make short term decisions while optimizing for a longer term goal. • Time plays a crucial role in Reinforcement problems. • There is no supervisor, only a real number or reward signal. • Goal: select action to maximize a future cumulative reward. Reinforcement Learning
8. An agent interacts with the environment, trying to take smart actions to maximize cumulative rewards. Image source Agent and Environment
9. Formulate a basic Reinforcement Learning problem 1. Environment (E) — Physical world in which the agent operates. 2. State (S)— Current situation of the agent. 3. Action (A) — All the possible moves that the agent can take. 4. Reward (R)— Positive Feedback from the environment. 5. Policy (π)— Method to map agent’s state to actions. It is a strategy which applies by the agent to decide the next action based on the current state. 6. Value (V) — Future reward that an agent would receive by taking an action in a particular state.
10. Guess the elements!! 🤔
11. Open-source Libraries for Implementation 1. Google TF-Agents - TF-Agents is a library for Reinforcement Learning in TensorFlow. 2. OpenAI Baselines - High-quality implementations of reinforcement learning algorithms. 3. Berkeley Ray RLLib - Reinforcement learning that offers both high scalability and a unified API for a variety of applications. 4. TensorForce - A TensorFlow library for applied reinforcement learning. 5. Stable Baseline - PyTorch version of Stable Baselines, reliable implementations of reinforcement learning algorithms.
12. Hands-on Colab-Notebook: https://bit.ly/3gVVOT4
13. Real-time Application
14. Games
15. RL in NLP • Summarization • Question answering • Knowledge-based QA • Machine Translation • Text generation Aligning Language Models to Follow Instructions
16. Autonomous Driving • AWS DeepRacer - the fastest way to get rolling with machine learning : Autonomous racing car that has been designed to test out RL in physical track. It uses cameras to visualize the runway and a reinforcement learning model to control the throttle and direction.
17. Recommendation
18. Trading and Finance Source
19. Industrial Automation
20. RL Resources 1. Ebook: Reinforcement Learning-An Introduction, a book by the father of RL - Richard Sutton : here 2. Lecture: Deep Reinforcement Learning: CS 285 at UC Berkeley 3. Course: Introduction to Reinforcement Learning with David Silver (From DeepMind’s research lead) 4. Course: Reinforcement Learning Specialization 5. Tutorial: Reinforcement Learning in 3 Hours | Full Course using Python 6. Open-Source repo: GitHub - yandexdataschool/Practical_RL: A course in reinforcement learning in the wild
21. Keep Exploring and Learning RL Agents!! You can find me at: www.linkedin.com/in/krupagaliya twitter.com/krupagaliya