Lade Inhalt...

Review on use of Reinforcement Learning in Artificial Intelligence

von Mehdi Samieiyeganeh (Autor) Parisa Bahraminikoo (Autor) G. Praveen Babu (Autor)

Forschungsarbeit 2012 5 Seiten

Informatik - Künstliche Intelligenz


Abstract - With the start of the 21st century, human moved into a new world of mechanics. Human started making machinery that can do the job for them. The technology developed so much that it started involving many other branches of engineering such as electronics, robotics etc. This eventually led to much more complex and smart machinery involving Artificial Intelligence.

Reinforcement Learning is a type of Machine Learning, and thereby also a branch of Artificial Intelligence. It allows machines and software agents to automatically determine the ideal behavior within a specific context, in order to maximize its performance . Reinforcement Learning (RL) comes from the animal learning theory. RL does not need prior knowledge, it can autonomously get optional policy with the knowledge obtained by trial-and-error and continuously interact with dynamic environment.

As a matter of fact, Reinforcement Learning is defined by a specific type of problem, and all its solutions are classed as Reinforcement Learning algorithms. In the problem, an agent is supposed decide the best action to select based on its current state. When this step is repeated, the problem is known as a Markov Decision Process.

A Markov Decision Process is a discrete time stochastic control process. At each time step, the process is in some state s, and the decision maker may choose any action that is available in state‘s’. Markov Decision Process provides a mathematical framework for modeling decision-making in situations where outcomes are partly random and partly under the control of a decision maker.

Keywords - Artificial Intelligence, Reinforcement Learning

I. Introduction

Since ancient days, philosophers and mathematicians developed a formal reasoning. The study and work of the mathematician Alan Turing led to invent the programmable digital electronic computer [14]. As per Turing Law of Computation, any mathematical deduction could be simulated by shuffling symbols such as “0” and “1”. A group of researchers continued with the research in neurology and information theory and cybernetics to develop an electronic brain.

In the year 1956, Artificial Intelligence was established at Dartmuth College during a conference. The people who were present for this conference were Marvin Minsky, Allen Newell, John McCarthy, and Herbert Simon. The programs written by them were beyond belief such as computers solving word problems in Algebra, proving logical theorems and speaking English. Herbert Simon envisaged that in twenty years, machines will be able to work like humans. Marvin Minsky believed that the problem of creating Artificial Intelligence would be solved considerably.

In human society, learning is an essential component of intelligent behavior. However, each individual agent need not learn everything from scratch by its own discovery.

Reinforcement Learning is a type of Machine Learning [13], and thereby, also a branch of Artificial Intelligence. It allows machines and software agents to automatically determine the ideal behavior within a specific context, in order to maximize its performance. Simple reward feedback is required for the agent to learn its behavior; this is known as the reinforcement signal. Reinforcement Learning is defined not by characterizing learning methods, but by characterizing a learning problem.

As mentioned Reinforcement Learning is an important machine learning method, its learning technology is divided into three types: Non-supervised Learning, Supervised Learning and Reinforcement Learning. Reinforcement Learning is an online learning technology [19] which is different from supervised learning and non-supervised learning. The reinforcement signal provided by the environment in Reinforcement Learning is to make a kind of appraisal to the action quality of intelligent agent, but not tell intelligent agent how to generate the correct action.

Planning under uncertainty is fundamental to solving many important real-world problems, including applications in robotics, network routing, scheduling, and financial decision making.

The rest of this paper is organized as follows: section 2 briefly describes the features of Artificial Intelligence and section 3 explains Reinforcement Learning.



ISBN (eBook)
581 KB
Institution / Hochschule
Jawaharlal Nehru University
review reinforcement learning artificial intelligence



Titel: Review on use of Reinforcement Learning in Artificial Intelligence