Neural Network
A mathematical model loosely inspired by the brain, made up of layers of interconnected nodes that learn to recognize patterns in data. Neural networks are the core architecture behind most modern AI — from image recognition to language models.
What is a Neural Network?
A neural network is a computational structure built from layers of simple processing units called nodes (or neurons). Each node takes in numerical inputs, applies a weighted transformation, and passes the result to the next layer. During training, the network adjusts its weights by comparing its outputs against known correct answers — a process called backpropagation. Over millions of iterations, the network learns to recognize patterns in data without being explicitly told the rules.
The "deep" in deep learning refers to neural networks with many layers — sometimes hundreds. Each layer learns progressively more abstract features: early layers might detect edges in an image or common word patterns in text; later layers recognize objects, entities, or meaning.
How Neural Networks Learn
Training a neural network requires three things: data (labeled examples of inputs and correct outputs), compute (GPUs processing millions of matrix multiplications), and a loss function (a measure of how wrong the current outputs are). The training loop adjusts weights to minimize that loss. The result is a model whose weights encode statistical patterns from the training data.
Feedforward networks: Simple layer-by-layer processing, used for classification and regression
Convolutional networks (CNNs): Specialized for image and spatial data
Recurrent networks (RNNs): Process sequences — largely replaced by transformers for most tasks
Transformers: The dominant architecture for language, vision, and multimodal AI today
Neural Networks in Operations
Operations teams do not build or train neural networks — they use the outputs of models that contain them. When an AI agent reads a scanned delivery note, the underlying document recognition runs on a neural network trained to parse layout and text. When a classifier routes incoming supplier emails, a neural network learned the patterns that distinguish order confirmations from complaints. Understanding that neural networks learn from examples — not rules — explains why AI systems sometimes fail on inputs that look similar to a human but are statistically different from the training data.