Machine Learning

Machine learning is a subfield of artificial intelligence. The aim is to build a mathematical description or model of the data we have in order to be able to gain new understanding about the system or to predict its future behavior. Approaches can be divided in three categories:

• Supervised learning – observations are labeled, meaning that each observation in a dataset belongs to a known class. The aim is to predict this class of new observations, where it is unknown. Some algorithms: linear and logistic regression, decision trees, support vector machines, artificial neural networks.
• Unsupervised learning – data is unlabeled. The goal is to discover new underlying patterns with minimum of human supervision. Examples of algorithms are clustering, principal component analysis and association rules.
• Reinforcement learning – does not need labeled data. An agent exists in an environment in which it takes actions towards accomplishing a goal. For each action it can be positively or negatively rewarded. After repeating the same sequence of actions multiple times, it seeks to maximize the award and minimize the effort. Thus, it learns the optimal way to accomplish a task. Two categories of algorithms are model-free and model-based algorithms.