WDHT
WDHT copied to clipboard
The vectorLoss and MSE loss
Hi,
I found the loss function in your code did not consistent with the ones in your paper, such as the MSE loss term in the code is ((h_i-h_j)^2-w_i^Tw_j), while in the paper is ((h_i-h_j)^2-0.5*(1-w_i^Tw_j)),
And the vectorLoss ia also wrong.
I wanna to know why you utiliz different losses from your paper?