Question about depth regression

[pnpmpnp]

I was looking at the code of AnyNet and had a question.
when you do disparity_regression in AnyNet, you do it for "-cost", is there a reason for doing it this way ?

"pred_low_res = disparityregression2(-self.maxdisplist[scale]+1, self.maxdisplist[scale], stride=1)(F.softmax(-cost, dim=1))"