sivvy

4 ms·

1. 1.▲
   by sivvy 9y ago

   Why not use conda?
2. 2.▲
   by sivvy 9y ago

   ...is $1.5M.
3. 3.▲
   by sivvy 9y ago

   Seems like in that case you would train both models separately on different cost functions. By phrasing it as a layer I was expecting both the SVM and the DNN could be trained simultaneously.
4. 4.▲
   by sivvy 9y ago

   Could you explain in a bit more detail how you would integrate an SVM layer into a DNN? The kernel matrix depends on all samples, while at training time you would only have access to those in the minibatch.
5. 5.▲
   by sivvy 10y ago

   I think he is actually a she.
6. 6.▲
   by sivvy 10y ago

   The premise of the optimizations in this article don't always hold, unfortunately. If x is floating point, sum(reverse(x)) is unlikely to be equal to sum(x) for long enough x, and sorting x on beforehand will make this effect even more