Neural network

Definition

A neural network is a computational model inspired by the structure of the human brain. It is made up of layers of interconnected nodes, or neurons, that process data and recognise complex patterns. Neural networks are a foundation of modern artificial intelligence and machine learning, especially in applications involving images, speech, and natural language.

By simulating how biological neurons activate and transmit signals, neural networks can adapt to data, learn from experience, and improve their accuracy over time.

Advanced

At an advanced level, neural networks consist of an input layer, hidden layers, and an output layer. Each neuron applies a weighted calculation and an activation function before passing information forward. Training involves adjusting weights through techniques such as backpropagation and gradient descent.

Deep learning models extend neural networks by adding many hidden layers, allowing them to solve complex problems such as real-time translation or self-driving car vision systems. Specialised architectures include convolutional neural networks for images and recurrent neural networks for sequential data.

Why it matters

• Powers AI breakthroughs in vision, language, and decision-making.
• Handles large, complex datasets better than traditional models.
• Enables automation in industries such as healthcare, finance, and logistics.
• Provides businesses with new capabilities in prediction and personalisation.

Use cases

• Image recognition for security and medical diagnostics.
• Natural language processing for chatbots and voice assistants.
• Predictive analytics in finance and supply chain management.
• Autonomous driving systems interpret sensor and camera data.

Metrics

• Accuracy, precision, and recall of predictions.
• Loss values during training and validation.
• Training time and computational cost.
• Model scalability and generalisation performance.

Issues

• Requires large datasets and high computational resources.
• Risk of overfitting if models memorise instead of generalising.
• Limited transparency in decision-making creates trust issues.
• Vulnerable to adversarial attacks in critical applications.

Example

A hospital uses a convolutional neural network to analyse X-ray images. The model identifies early signs of pneumonia with high accuracy, supporting doctors in diagnosis and improving patient outcomes.