Deep Learning (DL)

Deep Learning (DL) is a subfield of machine learning based on artificial neural networks with many hidden layers. It enables the automatic extraction of complex features directly from raw data and is used especially with large, high-dimensional datasets. Typical architectures include Convolutional Neural Networks (CNNs) for image processing and Transformer models based on Vaswani et al.¹ for sequential data such as speech or text.^{1 2 3} The high performance of deep learning now underpins numerous AI applications. Modern generative AI, such as GPT‑4⁴, is based on deep networks that can generate text, images, or multimodal content.¹

The input layer receives raw data such as pixel values or numerical variables. In the hidden layers, artificial neurons transform the inputs through weighting and activation functions, learning increasingly abstract features. The output layer finally provides the result, such as a class assignment or a numerical value, and serves as the basis for error calculation during training.² This provides the methodological foundation for generative AI systems that produce content autonomously. On this basis, methods of natural language processing can also be understood, as described in the next section.