Hello to all my dear friends!
This post talks about the Reinforcement Learning (RL) in the field of Artificial Intelligence (AI).
There are three types of learning in AI:
- Unsupervised learning – It is used for discovery of new patterns. Example is clustering group of people based on their similarities.
- Supervised learning – It is used to identify patterns. In this computer is feeded a lot of data with labels example photos of dogs and cats and the computer learns. Then given a photo, it can identify which is dog and which is cat.
- Reinforcement learning – This is different from both of above. Here computer performs random action without any supervision and is either rewarded (+1) or punished (-1) based on the action. This way computer learns itself what is the correct action to do in which situation. Example is how the robots learn to walk.
Reinforcement learning is also present in us humans. We do some action and if that gives a pain, then we do not do the same action again. Similarly for actions giving happiness, joy.
It does not require large labeled dataset like supervised learning. It is also innovative. It can learn a completely new approach/new solutions. It is also resistance to any bias unlike humans.
Real Life Use Cases of RL
- Google AlphaGo – Playing games like Go
- Google uses RL for energy efficiency in data centers
- Decision Service uses RL for advertisement and has improved click through rate
- Trendyol uses it for email advertising
- Alibaba uses it for advertising, which customer will click which ad.