The term early stopping originates from the field of artificial intelligence, particularly in machine learning, and also plays a role in the context of big data and smart data. Early stopping is a method used to prevent unnecessarily long learning when training computer models, for example for image recognition or self-learning systems.
Imagine teaching a software programme to distinguish between cats and dogs in photos. The system learns from many examples. If you train it for too long, it remembers too many specific details (such as the colour of a particular dog) and makes poorer decisions with new, unfamiliar images. This is why you monitor how well the system performs with new examples. As soon as no more improvement is seen, Early Stopping stops the training - this way the model remains flexible and also makes better decisions on new data.
This method saves time and computing power and ensures that results can be better utilised later in everyday practice. Early stopping is therefore an important building block in making artificial intelligence ready for use and economically efficient.
Deutsch 
English (UK)