AI Dictionary › Fondamenti AI

Data normalization

Normalizzazione dei dati

Data normalization is the technique of rescaling the numeric variables of a dataset into a common range, typically between 0 and 1, so that no feature dominates the others just because it is expressed in larger numbers. A model comparing annual income, in thousands, with age, in a couple of dozen units, risks giving excessive weight to income purely because of its scale, not its real importance.

Definition

How it works

The most common method, called min-max scaling, transforms each value by subtracting the variable's minimum and dividing by the difference between maximum and minimum, producing a range exactly between 0 and 1. The transformation must be computed on the training data and then applied identically to the test data and to future data, to avoid letting information from the test set indirectly influence training. It is sensitive to extreme values: a single outlier can compress all other values into a very narrow range.

Applications

It is an almost mandatory data preparation step for algorithms sensitive to variable scale, such as neural networks, k-nearest neighbors and support vector machines, while tree-based models, such as random forest and gradient boosting, are largely indifferent to scale and do not need it. It is applied systematically in computer vision, where pixel values are normalized before being fed into a neural network.

History & etymology

The idea of bringing different variables to comparable scales belongs to classical statistics well before machine learning, and the term normalization, from the Latin norma, rule, describes the idea of bringing different quantities under a shared rule. With the spread of neural networks from the 1980s and 90s onward, normalizing input data became standard practice, essential for making the training process stable and fast.

Related terms

• Standardization
• Feature engineering
• Dataset
• Outlier
• Neural Network

More in Fondamenti AI

• LLM (Large Language Model)
• Token
• Context Window
• Temperature
• AI Hallucination
• Fine-tuning
• RAG (Retrieval Augmented Generation)
• Embedding

Put it into practice

• Real scenario: apply this concept with an instant AI verdict.
• Daily gym: short drills to train dictionary concepts.
• Playoff game: challenge other players, elimination-style.

📱 Download the Android app (beta) iOS coming soon

Say what you mean. Get what you need.

Grace Certified, the AI coach that trains and certifies your prompt engineering, by Agora Intelligence.