Hands-On Deep Learning with R

Predicting class likelihood with softmax

The softmax() function is needed when there is more than one target variable. Softmax will create probabilities that a particular set of input variables belongs to each class. After these results are calculated, they can be used to assign input rows to one of the possible target classes. Let's explore this activation function with a slightly different example:

1. We will start by defining the function:

softmax <- function(x) {exp(x) / sum(exp(x))}

2. Next, let's pass a vector of values to the function:

results <- softmax(c(2,3,6,9))results[1] 0.0008658387 0.0023535935 0.0472731888 0.9495073791

3. Let's confirm that the sum of these transformed values equals 1:

sum(results)[1] 1

We can see that this function ...