In softmax, why is there a shuffle?

Part 1 (2017)
1

The order of operations shouldn’t really matter, so why shuffle?

I found this code in a tutorial!

def softmax(X):
    e_x = T.exp(X - X.max(axis=1).dimshuffle(0, 'x'))
    return e_x / e_x.sum(axis=1).dimshuffle(0, 'x')
2

this seems to be a function from Theano, which has reached its end of life. im not sure if theres a reason to still use it. anyway, dimshuffle appears to be a function that switches around the order of multi dimensional arrays. this has nothing to do with randomly shuffling the entries of an array.

3

Thanks for the explanation of the shuffling. I’m just using theano because it’s part of lesson 6. I guess I could be doing the same exercise in tensorflow or torch or any number of things. I don’t think it hurts to follow along the course though.