Zero-Shot Learning in Modern NLP

https://joeddav.github.io/blog/2020/05/29/ZSL.html

Zero-Shot Learning in Modern NLP | Joe Davison Blog
A latent embedding approach. A common approach to zero shot learning in the computer vision setting is to use an existing featurizer to embed an image and any possible class names into their corresponding latent representations (e.g. Socher et al. 2013).They can then take some training set and use only a subset of the available labels to learn a linear projection to align the image and label …
joeddav.github.io

Publié

dans

par

Étiquettes :