Is There a Widening Migration to a Gause Network?

Traditional ne netwacks are a bit over confident in general. So what you see in our results is that the confidence will be closer to the accuracy on the test data. And the adversary attack success is also decreased by a factor, which depends then on the data sad and the method, but it is observable. We also looked, of course, into the witalso we plotted, or we returned, some of these images which were misclassified. For example, a dog in a certain lighning, which is then misclassified as a deer to im looking archive version. But i don't see, i guess, as rich a result st as you're describing. It just seems