2 ms·
Can i please ask why lim{x->-inf} softmax(x) = 1/k ?
by seydor 3y ago
Can i please ask why lim{x->-inf} softmax(x) = 1/k ?
- hawkice 3y agoIt's splitting the probability across all x_i equally, even when they're all massively negatively weighted. This change would have all the probability going into the null option, in that case, basically.