Closed nfoti closed 9 years ago
The conversion code for nu didn't have the + 2 + D (and vice versa), which is standard in the literature. This doesn't make gradient computations invalid, but it does cause headaches when setting hyperparameters from previous papers.
nu
+ 2 + D
Thanks!
The conversion code for
nu
didn't have the+ 2 + D
(and vice versa), which is standard in the literature. This doesn't make gradient computations invalid, but it does cause headaches when setting hyperparameters from previous papers.