Reinforcement Learning 101

https://towardsdatascience.com/reinforcement-learning-101-e24b50e1d292

1. What is Reinforcement Learning? How does it compare with other ML techniques?

Reinforcement Learning(RL) is a type of machine learning technique that enables an agent to learn in an interactive environment by trial and error using feedback from its own actions and experiences.

Reinforcement Learning 101Reinforcement Learning 101
Reinforcement Learning 101Reinforcement Learning 101

2. How to formulate a basic Reinforcement Learning problem?

Some key terms that describe the basic elements of an RL problem are:

  1. State — Current situation of the agent
  2. Reward — Feedback from the environment
  3. Policy — Method to map agent’s state to actions
  4. Value — Future reward that an agent would receive by taking an action in a particular state
Reinforcement Learning 101Reinforcement Learning 101
Reinforcement Learning 101Reinforcement Learning 101
<iframe allowfullscreen="allowfullscreen" class="es ff fb fl v" data-mce-fragment="1" frameborder="0" height="480" scrolling="auto" src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FQilHGSYbjDQ%3Ffeature%3Doembed&display_name=YouTube&url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DQilHGSYbjDQ&image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FQilHGSYbjDQ%2Fhqdefault.jpg&key=a19fcc184b9711e1b4764040d3dc5c07&type=text%2Fhtml&schema=youtube" title="Deep Reinforcement Learning in Pac-man" width="854"></iframe>

3. What are some of the most used Reinforcement Learning algorithms?

Q-learning and SARSA (State-Action-Reward-State-Action) are two commonly used model-free RL algorithms. They differ in terms of their exploration strategies while their exploitation strategies are similar. While Q-learning is an off-policy method in which the agent learns the value based on action a* derived from the another policy, SARSA is an on-policy method where it learns the value based on its current action derived from its current policy. These two methods are simple to implement but lack generality as they do not have the ability to estimates values for unseen states.

Reinforcement Learning 101Reinforcement Learning 101

4. What are the practical applications of Reinforcement Learning?

Since, RL requires a lot of data, therefore it is most applicable in domains where simulated data is readily available like gameplay, robotics.

  1. In robotics and industrial automation, RL is used to enable the robot to create an efficient adaptive control system for itself which learns from its own experience and behavior. DeepMind’s work on Deep Reinforcement Learning for Robotic Manipulation with Asynchronous Policy updates is a good example of the same. Watch this interesting demonstration video.
<iframe allowfullscreen="allowfullscreen" class="es ff fb fl v" data-mce-fragment="1" frameborder="0" height="480" scrolling="auto" src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FZhsEKTo7V04%3Fstart%3D48%26feature%3Doembed%26start%3D48&url=http%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DZhsEKTo7V04&image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FZhsEKTo7V04%2Fhqdefault.jpg&key=a19fcc184b9711e1b4764040d3dc5c07&type=text%2Fhtml&schema=youtube" title="Deep Reinforcement Learning for Robotic Manipulation" width="854"></iframe>

5. How can I get started with Reinforcement Learning?

For understanding the basic concepts of RL, one can refer to the following resources.

  1. Teaching material from David Silver including video lectures is a great introductory course on RL.
  2. Here’s another technical tutorial on RL by Pieter Abbeel and John Schulman (Open AI/ Berkeley AI Research Lab).
  1. DeepMind Lab is an open source 3D game-like platform created for agent-based AI research with rich simulated environments.
  2. Project Malmo is another AI experimentation platform for supporting fundamental research in AI.
  3. OpenAI gym is a toolkit for building and comparing reinforcement learning algorithms.
上一篇:PAT_A1014 Waiting in Line


下一篇:在 LINQ to Entities 查询中无法构造实体或复杂类型