• morte [she/her]@hexbear.net
    link
    fedilink
    English
    arrow-up
    2
    ·
    8 months ago

    What if we embed genders in hyperdimensional vectors and train a neural network to classify someone’s “unique” gender

    • Hmm, interesting idea, but what if the kid incorporates a xenogender axis orthogonal to all previously known xenogender axes? My algorithm would project away this axis and it would have no effect on the computed output, but a neural network would have randomly determined weights connected to the input corresponding to that axis, resulting in unexpected results