19:3219:32, 3 July 2022diffhist+901
Frontiers Media
I am not quoting the two facts independently and try to avoid a direct conclusion. I hope this is deemed more appropriate. If this is not received well, can we move this to the discussion page of the article?Tag: Reverted
13:3313:33, 18 October 2019diffhist+557
Deep learning
→Interpretations: Added a reference that shows that universal approximation also holds for ReLU. I deem this as important information, as ReLU is one of the most used activation functions in deep learning today. Before this work, it was not clear whether such layers would still be universal function approximatiors.
10 October 2019
12:1812:18, 10 October 2019diffhist+56
Pattern recognition
Fixed the dubious statement on pattern recognition focusing only on supervised methods. Many PR text books include unsupervised methods.Tags: Mobile editMobile web edit
19:5319:53, 7 October 2019diffhist−97
Projection-slice theorem
Added the Radon transform to the term "project" to clarify which kind of projections is referred to. This should solve one issue reported on the "talk" page.
22:1522:15, 6 October 2019diffhist+292
m
Deep learning
→Overview: Added a reference for the greedy construction of deep networks; Changed the qualifier from "often" to "can be". Reference has more than 4000 citations on google scholar and should be appropriate for wikipedia.
09:5009:50, 6 October 2019diffhist+1,678
Emission computed tomography
Added an introduction to emission tomography. Passage is a direct quote from the reference. Note that the reference is under CC 4.0 BY. Therefore, the passage can be reused here.