Augmented Intelligence Newsletter (AiN) # 9: Reinforcement Learning

Reinforcement Learning

Welcome to Augmented Intelligence Newsletter (AiN) by C. Naseeb.

AiN Issue # 9

Thank you for reading my latest article on?Understanding Reinforcement Learning, its key concepts and applications.

Here at?LinkedIn?and at Medium?I regularly write about business, technology, digital transformation, and emerging trends. To read my future?articles simply?subscribe to this newsletter or click 'Follow'.?

Hey, in this issue, I explain the key concepts around RL: Understanding Reinforcement Learning, its key concepts and applications.

Reinforcement Learning

Learning is defined as acquiring knowledge or skills through experience, study. When it comes to Machine Learning, it is about empowering computers to learn without explicit codified instructions. It is the study of computer algorithms that can improve automatically through experience (learning) and data. For more details, look at the previous newsletter, in which we elaborated on the understanding of AI, ML, DL, and Data Science.?

Machine learning algorithms build a model based on sample data, known as training data, to make predictions or decisions without explicitly codified instructions. Some applications of Machine learning algorithms include but are not limited to medicine, email filtering, speech recognition, and computer vision; where it is difficult or unfeasible to develop conventional algorithms to perform the needed tasks. In this article, we will concentrate on?Reinforcement Learning.?

It is a machine learning method that deals with how software agents or machines should take actions in an environment. The agents are trained with a reward and punishment mechanism; they get rewarded for correct moves and punished for bad ones. The goal for the agent is to minimize the wrong moves and maximize the right ones so that it gets the maximum reward.?

Reinforcement Learning is part of the Deep Learning method, which can help you achieve a difficult objective or maximize a specific dimension over many iterations. Let's discuss some important terms to understand it further, before going deep into the applications of RL.?

Agent. An entity that performs some action(s) in the?Environment?to get some?Reward.?

Environment.?A scenario that an agent has to face, or a situation in which an agent can operate.?

Reward. A return given to an agent when it acts.?

Policy. The agent's strategy to take the next action based on its current state.?

State. It refers to the current situation returned by the Environment.?

Value. It is the long-term return as compared to short term reward.?

Model of the Environment: A representation of the behavior of the environment. It helps to make inferences about the environment's behavior.?

Q Value or Action Value. It is similar to the value, with only one difference that it also takes the current action as an additional parameter.?

Applications of Reinforcement Learning

There are potentially many application areas for Reinforcement Learning to have its say, and some of them include Healthcare, Self-driving cars, Trading & Finance, Intelligent Automation, Engineering, Gaming, Natural language Processing (NLP), Marketing and advertising, Robotics, and many more. Here, we will discuss the applications of RL in specific scenarios to a limited extent.?

In an industrial automation scenario for the application of reinforcement, learning-based robots are used to execute various tasks. Apart from the point that these robots are more efficient than human beings, they can also conduct tasks that would be treacherous for people.?

Deepmind uses Reinforcement Learning AI agents to cool Google Data Centers. This led to a 40% reduction in energy?consumption. The centers are now completely managed with the AI system without human intervention. There is still supervision from data center experts. The system works in the following way:

• It takes snapshots of data from the data centers every five minutes and feeds them to deep neural networks
• It then predicts how distinct mixtures will affect future energy spendings
• It identifies actions that will lead to minimal power consumption while maintaining a set standard of safety criteria?
• Dispatching and implementing these actions at the data center
• The local control system verifies the actions.?

Each of the applications of the RL demands a separate article to discuss the details. However, hopefully this article has given you a good grasp on the understanding the fundamental concepts, and where it can be used.

Subscribe and view previous issues?here.?

Enjoy the newsletter! Help us make it great and better by sharing it with your network.?

Have a nice day! See you soon. - Chan