Hadamard codification in computer vision

Angello Hoyos (CIMAT)

One-hot class labeling for supervised training has been widely used in machine learning due to its simplicity. However, more questions need to be asked about whether this is the best alternative. This talk will discuss Hadamard codes as an alternative for classic computer vision problems, classifying adversarial examples, and image segmentation in applications such as autonomous vehicles, medical images, among others.