{"id":40720,"date":"2025-10-22T15:43:50","date_gmt":"2025-10-22T13:43:50","guid":{"rendered":"https:\/\/csw.agency\/glossar\/deep-learning\/"},"modified":"2025-10-22T15:45:47","modified_gmt":"2025-10-22T13:45:47","slug":"deep-learning","status":"publish","type":"glossar","link":"https:\/\/csw.agency\/en\/glossar\/deep-learning\/","title":{"rendered":"Deep learning"},"content":{"rendered":"

Deep learning is a specialized sub-area of the field of computer science. machine learning<\/a> (machine learning) and thus the Artificial intelligence (AI)<\/a>. This technology enables computer systems to learn independently and without explicit programming from huge amounts of data and to recognize complex patterns. Deep learning differs from traditional machine learning methods in that it uses hierarchical neural networks that comprise several processing layers - so-called \u201ehidden layers\u201c.<\/p>\n

Structure and mode of operation<\/h2>\n

Deep learning is based on artificial neural networks (ANNs), which are loosely based on the structure and functioning of the human brain. A typical neural network consists of an input layer, several hidden layers and an output layer. While conventional neural networks often only have a few hidden layers, deep learning networks are characterized by a high number of these layers, sometimes up to 150 or more. The term \u201edeep\u201c refers directly to this multi-layered nature.<\/p>\n

Each layer in the network specializes in extracting features with different levels of abstraction from the input data. For example, the first layer could recognize simple features such as edges or colors in an image, while deeper layers identify more complex structures such as shapes, textures or even entire objects. This process of automatic feature extraction is a significant advantage over traditional machine learning, where features often have to be defined manually by experts.<\/p>\n

Training a deep learning model typically requires very large amounts of high-quality data and considerable computing power, often supported by graphics processing units (GPUs). Using algorithms such as backpropagation, the network continuously adjusts its internal parameters (weights and biases) to improve the accuracy of its predictions or classifications. Common frameworks for the implementation of deep learning applications are TensorFlow, PyTorch or Keras.<\/p>\n

Areas of application and significance<\/h2>\n

Deep learning has undergone rapid development in recent years and is now a key technology in numerous areas. Its ability to recognize complex patterns in unstructured data such as images, videos and texts enables groundbreaking applications:<\/p>\n