What is the role of softmax in a model and what is it used for?

Softmax is an activation function that transforms a vector of arbitrary numbers into a vector of probabilities, which sum to 1. In machine learning models, especially in classification tasks, softmax is used at the output layer of a neural network to obtain the probabilities of belonging to each class. This allows interpreting the model's output as a probability distribution and selecting the class with the highest probability.

Example: if the model outputs values [2.0, 1.0, 0.1], applying softmax transforms them into probabilities approximately [0.65, 0.24, 0.11].