What is a consequence of not normalizing data in machine learning?

Not normalizing data in machine learning can lead to reduced model accuracy because many algorithms rely on the scale of the data to make predictions or classifications effectively. For instance, algorithms like K-nearest neighbors (KNN) and gradient descent algorithms, which rely on distance calculations, can be heavily influenced by the scale of the input features. If features are on different scales, the model may give more weight to larger values, leading to a distorted representation of the data.

Moreover, if features have vastly different ranges, the model may struggle to learn effectively because it could converge to a local minimum that is not optimal, resulting in poor performance. Normalization helps to ensure that every feature contributes equally to the distance calculations and subsequently to the model's performance. Therefore, not normalizing can lead to inconsistencies in how data points are treated, ultimately diminishing the model's accuracy and reliability.