Contrastive learning is a term from the fields of Artificial Intelligence, Big Data and Smart Data, and Digital Transformation. It describes a specific method for computers to learn to better understand and differentiate between data.
Imagine you want to teach software to recognise the difference between dogs and cats. In contrastive learning, you show the computer many images: some where the animals look similar, and others where they look very different. The software then learns which features are typical of dogs and which are typical of cats by examining the differences („contrasts“) between the images very closely.
Contrastive learning is particularly helpful when there is little information available or when it is difficult to label everything manually. Modern artificial intelligence can use this technique to find patterns, for example, to detect unusual transactions in large amounts of financial data (such as in banks) or to monitor machines in a factory. This saves time, makes processes safer, and helps to make smarter decisions.
In short: Contrastive learning is an innovative learning method for computers to better recognise differences in data and derive practically usable insights from them.
Deutsch 
English (UK)