Function dfdx::data::one_hot_encode

pub fn one_hot_encode<const B: usize, const N: usize>(
    class_labels: &[usize; B]
) -> Tensor2D<B, N>

One hot encodes an array of class labels into a Tensor2D of probability vectors. This can be used in tandem with crate::losses::cross_entropy_with_logits_loss().

Const Generic Arguments:

• B - the batch size
• N - the number of classes

Arguments:

• class_labels - an array of size B where each element is the class label

Outputs: Tensor2D with shape (B, N)

Examples:

let class_labels = [0, 1, 2, 1, 1];
// NOTE: 5 is the batch size, 3 is the number of classes
let probs = one_hot_encode::<5, 3>(&class_labels);
assert_eq!(probs.data(), &[
    [1.0, 0.0, 0.0],
    [0.0, 1.0, 0.0],
    [0.0, 0.0, 1.0],
    [0.0, 1.0, 0.0],
    [0.0, 1.0, 0.0],
]);