4 ms·
Right, so the algorithm that trains the neural net can be explained (autodifferentiation, backprop etc) but the neural net's functioning itself is billions of
by habitue 3y ago
Right, so the algorithm that trains the neural net can be explained (autodifferentiation, backprop etc) but the neural net's functioning itself is billions of parameters which we absolutely cannot fully explain how it makes decisions