Generalized Zero-Shot Learning using Identifiable Variational Autoencoders

作者：

Highlights：

• Identifiable VAE is a generative model to address conventional and generalized ZSL.

• To learn a latent space that approximates the actual data distribution.

• Extensive experiments on CUB, AWA1, AWA2, SUN and aPY datasets are performed.

摘要

•Identifiable VAE is a generative model to address conventional and generalized ZSL.•To learn a latent space that approximates the actual data distribution.•Extensive experiments on CUB, AWA1, AWA2, SUN and aPY datasets are performed.

论文关键词：Zero-shot learning,Generalized zero-shot learning,Non-Linear ICA,Disentangled Representation Learning

论文评审过程：Received 5 April 2021, Revised 19 September 2021, Accepted 19 November 2021, Available online 2 December 2021, Version of Record 12 December 2021.

论文官网地址：https://doi.org/10.1016/j.eswa.2021.116268