Confusion in paper #17
-
|
Hi, thought it was possible to ask here. In chapter 4.4 Comparing DL models you compare first the models on the datasets without using an ensemble structure for ResNet and FT-Transformer. NODE has a significantly high performance on a few datasets. Then you compare the three all using ensemble structures and say that the results indicate that FT-Transformer and ResNet benefit more from ensembling, but then say that the gap between ResNet and NODE is significantly reduced with ensembling. I would think the gap has increased between NODE and ResNet based on the wording. |
Beta Was this translation helpful? Give feedback.
Replies: 0 comments 2 replies
-
|
Hi! By "the gap is reduced" we mean "the gap became smaller". Which is in line with the statement "FT-Transformer and ResNet benefit more from ensembling", since "more benefit" leads to "smaller gaps" on the problems where NODE is superior to ResNet. |
Beta Was this translation helpful? Give feedback.
-
|
Yes, thanks |
Beta Was this translation helpful? Give feedback.
Hi! By "the gap is reduced" we mean "the gap became smaller". Which is in line with the statement "FT-Transformer and ResNet benefit more from ensembling", since "more benefit" leads to "smaller gaps" on the problems where NODE is superior to ResNet.
Does this help?