Machine Learning (ML)

Machine Learning (ML) - German: maschinelles Lernen - is a central subdiscipline of Artificial Intelligence. The goal is to develop systems that learn from data to recognize patterns and derive predictions or decisions. Unlike knowledge-based approaches where rules must be explicitly programmed, an ML model learns these automatically by analyzing large datasets.¹

The development of an ML model takes place through training on a large data basis. Algorithms are developed that extract rules from the data and store them in a model. This model can then be applied to new, previously unseen data.²

Types of Machine Learning

Below, four types of machine learning are presented, illustrated in the figure below according to Schneider (2025)³: (1) supervised learning, (2) unsupervised learning, (3) semi-supervised learning, and (4) reinforcement learning.

Supervised Learning

Here, the model is presented with pairs of inputs and their corresponding expected outputs (labels). The goal is to learn a function that classifies or predicts new input data as accurately as possible.^{2 3}

Unsupervised Learning

In this case, the model receives only input data without associated target values. It tries to identify hidden structures or patterns, for example through clustering or dimensionality reduction.^{2 4}

Semi-Supervised Learning

This learning form combines supervised and unsupervised learning. It is used especially when only a small portion of the available data is labeled while most is unlabeled. In such scenarios, deep learning methods are used to generate so-called pseudo-labels from the unlabeled data. The goal is to classify input data as accurately as possible.^{2 5}

Reinforcement Learning

This approach is based on the interaction of an agent with an environment. The agent makes decisions (actions) and receives rewards or penalties depending on the outcome. Through this feedback, it learns a strategy to optimize its actions - a so-called optimal policy.^{1 2 6}