The term "learning representation" is particularly at home in the fields of artificial intelligence, big data, smart data and automation. It refers to how a machine - for example a computer programme - represents data in such a way that it can learn from it. The machine attempts to recognise important patterns and correlations in the data independently.
Imagine you want to teach a computer to recognise different types of fruit from pictures. The "Learning Representation" ensures that the program recognises: a banana is crooked and yellow, an apple is round and red or green. The programme processes the images in such a way that these differences stand out clearly.
The better this visualisation, the easier it is for the system to "learn" - for example, to correctly assign new photos without having seen every image before. This is important because with big data, huge amounts of data have to be utilised sensibly in a short space of time.
Learning representation therefore plays a key role in gaining knowledge from data and making automated decisions. It forms the basis of many modern applications, such as speech or image recognition, process automation or recommendations in online shopping.
Deutsch 
English (UK)