HIGHLIGHTS
- who: Tiago de Souza Farias from the Northeast Normal University, China have published the Article: Feature alignment as a generative process, in the Journal: (JOURNAL)
- how: The goal in this paper is to show that feature alignment can be used for approximate reversibility of neural_networks as well as sampling of images. The main difference is that in these previous articles the features are obtained with a pretrained network. Equation is similar to activation maximization except that the authors are minimizing for the neurons to have a target frontiersin.org 10.3389/frai.2022.1025148 This . . .
If you want to have access to all the content you need to log in!
Thanks :)
If you don't have an account, you can create one here.