One Hot Encoding

# Let `classes` be string classes
cat, cat_inv = np.unique(classes, return_inverse=True)
one_hot = np.zeros((classes.size, cat.size))
one_hot[np.arange(classes.size), cat_inv] = 1