Using Artificial Intelligence That Deals With Training An Agent

1876 Words Sep 11th, 2016 8 Pages
Machine learning is the subfield of artificial intelligence that deals with training an agent to perform a task or a set of tasks, without the need to be explicitly programmed. The learning process is made possible as the agent uses previously observed data to make decisions, while at the same time, it receives feedback for its actions and as a result, it is able to improve its performance on future tasks.
Different subcategories of machine learning algorithms are defined according to the way the agent receives its feedback. The three main subcategories are: supervised learning, unsupervised learning and reinforcement learning [30], however, this distinction is not absolute, as some algorithms have characteristics from more than one of these categories.
In supervised learning, the algorithm is trained on input-output examples, called labeled data, and its goal is to find a function that effectively maps new inputs to the desired labels. This category of algorithms is usually used for classification and regression tasks.
In unsupervised learning, the algorithm tries to discover patterns in the input data, without being given any specific instructions about the output. Such algorithms are used to perform clustering and dimensionality reduction tasks. Lastly, in reinforcement learning, the agent learns a policy on how to act, given a specific state of its environment, by receiving either a punishment or a reward for its actions by the environment. This way, it can alter its…

Related Documents