What is a neural network?
A neural network is a computational model loosely inspired by the human brain, consisting of interconnected layers of computing nodes that learn to recognise patterns in data.
DEFINITION
A neural network consists of layers of computing nodes, called neurons, that are connected to each other. As data passes through the network, each layer computes a transformation of the input and passes the result to the next layer. During training, the network adjusts the strength of its connections until it can predict the desired outputs as accurately as possible. Neural networks such as Convolutional Neural Networks (CNNs) recognise images, recurrent networks (RNNs) process sequences, and Transformer architectures are the foundation of modern language models. Deep learning refers to neural networks with many layers. Behind AI successes in image recognition, language processing, or games, there is almost always a neural network. The parallel to the human brain is a simplification: neural networks work algorithmically, not biologically.
CONNECTIONS
Leadership
Leaders do not need to program neural networks, but they should understand how learning processes work. Those who grasp the logic of training and improvement through data can assess and steer AI projects more effectively.
Agility
When teams develop AI systems with neural networks, they need clear acceptance criteria. A Definition of Done for model quality is just as important as one for software features.
Project Management
Neural network projects have high unpredictability in outcomes. This model risk must be explicitly recorded in the risk register, including fallback plans for insufficient model quality.
KEY POINTS
- Neural networks learn by adjusting connection weights during training.
- Deep learning refers to networks with many layers.
- Transformer networks are the foundation of modern language models such as GPT.
- Neural networks underpin image recognition, language processing, and more.
- The parallel to the brain is conceptual, not biologically precise.
EXAMPLE
You upload a photo to an app, and the app immediately recognises which friends are in it. Behind this function is a Convolutional Neural Network trained on millions of face images. It has learned to recognise eyes, noses, and distinctive facial features as patterns and combine this information layer by layer.
MISCONCEPTIONS
Is a neural network the same as a brain?
No. Neural networks are inspired by brains but are built algorithmically in a completely different way. There is no consciousness, no biological signal processes, and no direct structural correspondence.
Are deeper neural networks always better?
Not automatically. Deeper networks can learn more complex patterns but need more data, more computing power, and are harder to train. The right depth depends on the use case.