Entrée normalisée is a crucial preprocessing step in the fields of intelligence artificielle (AI) and apprentissage automatique. It involves adjusting the input data values to a common scale without distorting differences in the ranges of values. This is particularly important when dealing with datasets that contain features with varying units and scales, as it helps improve the performance and convergence speed of algorithms.
Techniques de normalisation typically transform the input data to a standard range, such as [0, 1] or a mean of 0 and a standard deviation of 1. Common methods include:
- Normalisation Min-Max: This technique rescales the feature to a fixed range, usually [0, 1]. The formula is:
X' = (X - min(X)) / (max(X) - min(X)). - Normalisation Z-score : This method standardizes the features by removing the mean and scaling to unit variance, using the formula:
X' = (X - μ) / σ, where μ is the mean and σ is the standard deviation. - Échelle décimale : This involves moving the decimal point of values of the feature. The number of decimal points moved depends on the maximum absolute value of the feature.
La normalisation des données d'entrée est essentielle pour diverses les applications d'IA, particularly in neural networks, where the fonctions d'activation can be sensitive to the input data scale. By ensuring that the input features are on a similar scale, normalized inputs help in reducing bias during the training process, facilitating better learning and more accurate predictions.