What are neural networks?
Neural networks are computer-based models that mimic the way the human brain works and represent a core technology in AI and machine learning. They process data through weighted connections between artificial neurons, identify patterns in language, images, text, or numbers, and enable data-driven learning for decision support and forecasting.
| Feature | Details |
| Category | AI / machine learning / algorithms |
| Areas of application | Speech and image recognition, forecasting, fraud detection, business intelligence, text analysis |
| Typical use cases | Pattern recognition, classification, predictive models, segmentation, anomaly detection |
| Related terms | Artificial intelligence, deep learning, machine learning, backpropagation, convolutional neural network |
| Benefits | Automated pattern recognition, process optimization, accurate forecasting, decision support |
At a glance
Process data through weighted connections between neurons.
Enable decision support through data-driven learning.
Identify patterns in language, images, text, or numbers.
Provide the technological foundation for deep learning and modern AI applications.
Different types of neural networks vary depending on their structure and intended use.
Definition: neural networks explained simply
Neural networks are computer-based models inspired by the way the human brain works — which is why they are often referred to as artificial neural networks. As a field of artificial intelligence (AI), neural networks form the foundation for many modern machine learning methods. They consist of multiple layers of interconnected artificial neurons that can process information, identify patterns, and make predictions based on them. By training on large datasets, neural networks continuously improve their results.
This makes it possible to automate complex tasks such as speech and image recognition, fraud detection, and forecasting — often in a fraction of the time it would take a human to perform them. Companies use neural networks to optimize processes, analyze large volumes of data, and support intelligent decision-making.
Neural networks and AI: are neural networks artificial intelligence?
Neural networks are a core tool within artificial intelligence, but they are not synonymous with AI. Rather, they form the foundation of many modern AI applications. While AI is the broad term for methods that enable machines to exhibit intelligent behavior, neural networks are specialized algorithms that enable this behavior by learning from data.
| Term | Classification |
| Artificial intelligence (AI) | Umbrella term for methods that enable machines to exhibit intelligent behavior |
| Machine learning | Subfield of AI: systems learn from data without being explicitly programmed |
| Neural networks | Specialized algorithms within machine learning, inspired by the human brain |
| Deep learning | A type of neural network with many layers for particularly complex tasks |
In short: Neural networks are the engine behind the “intelligence” of many modern AI applications. They are a subfield of machine learning and therefore an essential, but not the only, component of AI.
Neural networks vs. deep learning: are neural networks deep learning?
| Feature | Neural networks | Deep learning |
| Layers | Few hidden layers | Many deeply nested hidden layers |
| Complexity | Low to moderate | High |
| Data requirements | Moderate | Very large |
| Computational effort | Low to moderate | High |
| Typical tasks | Classification, simple patterns | Language, images, complex forecasting |
Neural networks and deep learning are often discussed in the same context, but they differ in terms of complexity and application. Neural networks are the basic structure, while deep learning refers to neural networks with many hidden layers that are particularly deep.
These deeper structures enable deep learning models to solve highly complex tasks. While simple neural networks can already recognize patterns, deep learning goes a step further by using large amounts of data and substantial computing power to deliver highly accurate results. In practice, every deep learning model is a neural network — but not every neural network is automatically deep learning.
How do artificial neural networks learn?
Artificial neural networks learn by analyzing example data and deriving patterns and relationships from it. Their structure is inspired by the human brain: They consist of many interconnected nodes — the artificial neurons — that receive, process, and pass on information. A neural network is divided into an input layer, one or more hidden layers, and an output layer. The input layer receives the raw data. The hidden layers identify features and calculate intermediate results, while the output layer produces the final prediction or classification.
The learning process at a glance:
-
- The network processes input data together with the corresponding correct results (labels)
- Each layer identifies specific features and passes information onward
- The calculated output is compared with the actual result
- The difference (error value / loss) is calculated
- The weights between neurons are adjusted through backpropagation
- This process is repeated thousands of times until optimal weights are found
The learning process used by neural networks is usually based on supervised learning. In this approach, the network is provided with input data together with the corresponding correct results (labels). First, the network processes the data layer by layer: Each layer identifies specific features and passes the information to the next layer. At the end, the network compares its calculated output with the actual result. The difference between the prediction and reality is referred to as the error value (loss).
To minimize this error, the neural network automatically adjusts the connections between neurons, known as weights. This adjustment is performed using learning algorithms such as backpropagation in combination with optimization methods such as gradient descent. The process is repeated thousands of times until the network has found the best weights. Through this iterative process, prediction accuracy gradually improves.
Thanks to this learning method, artificial neural networks can independently analyze previously unseen data, identify patterns, and make accurate predictions. For example, they can understand speech, recognize images, or support forecasting in business intelligence.
What types of neural networks are there?
Neural networks can be divided into different types that vary in structure and application. The most important types of artificial neural networks include:
Perceptron is the simplest form of a neural network and is considered the foundation of modern network architectures. It typically consists of a single artificial neuron that assigns weights to input values, sums them, and makes a binary decision using an activation function.
Feed-forward neural networks (FFNNs) are the standard for many applications. Information flows in only one direction: from the input layer through one or more hidden layers to the output layer. There are no feedback loops — hence the name “feed-forward.” When particularly many layers are used, the network is also referred to as a deep feed-forward neural network.
Convolutional neural networks (CNNs) are specifically designed to work with image or video data. They can extract important features from two- or three-dimensional input data. This makes CNNs particularly well suited to object recognition, image recognition, and classification.
Recurrent neural networks (RNNs) extend conventional neural networks with feedback connections, allowing information to flow not only forward but also backward through the network. This gives these networks a kind of memory, making them ideal for tasks where context matters — for example, speech processing, text analysis, or video.
Bissantz and neural networks
At Bissantz, neural networks are used with the same principle that applies to all of the company’s technologies: They should serve a practical purpose, not be technology for technology’s sake. In the DeltaMaster business intelligence software, neural network methods are used to automatically identify data patterns, filter out relevant information, and support decision-makers with clear, easy-to-understand visualizations — for example, when prioritizing key figures or performing intelligent segmentation.
The goal is not to use black-box models without critical reflection, but to combine the capabilities of neural networks with the explainable logic of business management and control. For example, a neural network may identify an unusual correlation in a dataset. At Bissantz, this insight is presented in a context that makes it understandable and actionable for business users.
FAQ: frequently asked questions
A neural network is a computer program that works somewhat like the human brain: It consists of many interconnected “nodes” that pass on and process information. Through training with example data, it learns to recognize patterns — for example, whether an image shows a cat or whether a series of numbers indicates a decline in sales.
A conventional algorithm follows fixed, predefined rules. A neural network, by contrast, learns its own rules from data — it is not explicitly programmed but trained. This makes it more flexible for complex, unstructured tasks, but also less directly transparent.
No, at least not in technical detail. Modern BI solutions such as DeltaMaster apply neural networks. What matters is understanding what the results mean and how they feed into decisions, not how the model performs its calculations internally.
Yes. Neural networks are statistical models and can make mistakes — particularly when dealing with unfamiliar data situations, biased training data, or insufficient training. This is why principles such as explainable AI and human-in-the-loop are essential for critically reviewing AI results and keeping responsibility with humans.
DeltaMaster uses neural network methods for automated pattern recognition, intelligent prioritization of key figures, and context-based analysis of anomalies. The results are always presented transparently, making them understandable to business users without requiring technical expertise.
Summary
Neural networks are one of the fundamental technologies behind modern artificial intelligence: They enable machines to learn from data, recognize patterns, and make accurate predictions — from image recognition to business forecasting. In a business context, they deliver the greatest value when their results are transparent, explainable, and embedded in clear, easy-to-understand visualizations. Bissantz uses neural networks in DeltaMaster selectively and responsibly — as a powerful tool for intelligent, explainable business management and control.
Related terms: deepen your knowledge
Free of charge for you
How AI reduces the effort required for interpretation—and helps companies move
more quickly from analysis to action
