Backpropagation is a fundamental concept in Artificial Intelligence and is frequently used in the fields of Big Data and Smart Data, as well as digital transformation. It is a method by which artificial neural networks „learn“, meaning they improve their results step by step.
Imagine an intelligent system is meant to recognise whether a photo depicts a dog or a cat. Initially, the system makes a lot of mistakes because it doesn't yet know what a dog or a cat looks like. With backpropagation, the system can check after each attempt how far off it was, and adjust its internal settings (its „dials“, so to speak) so that it can make a better decision next time.
Backpropagation thus works like a learning process: when an error is detected, it's traced back to where in the system the error originated, and adjustments are made there. This is repeated many times with many examples, so that the system gets better and better.
This learning method is the key to enabling today's AI applications, such as speech recognition, image recognition, and automatic translations, to function at all. Backpropagation ensures that intelligent systems learn from their mistakes.
Deutsch 
English (UK)