3 ms·
11D tensor? Where would that get us with TensorFlow? To hierarchical attention?
by bitL 7y ago
11D tensor? Where would that get us with TensorFlow? To hierarchical attention?
- DoctorOetker 7y agoHahaha! Surely you must be trolling ;) spoiler alert: in math and physics, the dimension of a tensor is simply the dimension of your space, i.e. the number of basis vectors, and the number of indices is called the order (also degree or rank) while datamonkeys on MatrixBurp or TensorFart call the latter the "dimension" and the former ... the ... euh depth? So just like many here are objecting on the usage of "dimension" in the submitted article, (s)he is trolling the very same practice many ML people do with respect to mathematicians and physicists regarding the "dimension" of a tensor. so in math or physics terminology an 11-dimensional rank 2 tensor would take up just 11 ^ 2 = 121 values, while a 2-dimensional rank 11 tensor would take up 2 ^ 11 = 2048 values.