Neural networks

Neural networks, more fully artificial neural networks, are computing systems built from many simple processing units working in parallel, loosely inspired by the neurons of the brain. Each unit holds a fragment of learned knowledge plus rules about relationships, and the network as a whole learns to recognize patterns in large, noisy datasets. Food companies apply them to model processes and predict how foods behave under given conditions: forecasting shelf life, tuning drying and baking curves, spotting defects in machine-vision inspection, and predicting sensory scores from instrumental data. They complement mechanistic models wherever food matrices are too complex for first-principles equations, provided they are trained on sound experimental data.