Universal activation function for machine learning
Abstract This article proposes a universal activation function (UAF) that achieves near optimal performance in quantification, classification, and reinforcement learning (RL) problems. For any given problem, the gradient descent algorithms are able to evolve the UAF to a suitable activation function...
Enregistré dans:
| Auteurs principaux: | , , , |
|---|---|
| Format: | article |
| Langue: | EN |
| Publié: |
Nature Portfolio
2021
|
| Sujets: | |
| Accès en ligne: | https://doaj.org/article/761a26ef959a4ff2b37e7710b3ce6b10 |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
Soyez le premier à ajouter un commentaire!