Closed Luisaply closed 5 years ago
In 'utils.py' you have used a gaussian filter to get the auc.
z = ndimage.filters.gaussian_filter(z, 14)
Is there a reason to set sigma to 14?
If you see the closed issue, the original variance we use is 70. 14 is one-fifth of 70, and corresponds to the input size of the network.
In 'utils.py' you have used a gaussian filter to get the auc.
Is there a reason to set sigma to 14?