Why do you need to normalize your data in Machine Learning?

What is normalization?

Normalization is a process where values shifted between 0 to 1 range.

Why normalization is needed?

• If one feature in the dataset is larger in scale than the other, then this large-sized feature dominates during predictions.
• For NN, If the values are too high, the calculation takes a lot of time as well as memory. The same thing happens during backpropagation. As a result, the model converges slowly if the inputs are not generalized.

Where Normalization is very important?

• K-Means
• K-Nearest-Neighbours
• Principal Component Analysis (PCA)
• Gradient Descent

When Should You Use Normalization?

• If we don't know the data distribution
• If the distribution is not Gaussian/bell curve
• When your algorithm does not make assumptions about the distribution of your data, such as k-nearest neighbors and artificial neural networks.