Multimodal sensor fusion in the latent representation space

HIGHLIGHTS

SUMMARY

JMVAE (Joint Multimodal Variational Autoencoder)9 makes use of a joint inference network to learn the interaction between two modalities and they address the issue of missing modality by training an individual (unimodal) inference network for each modality as well as a bimodal inference network to learn the joint posterior, based on the product-of-experts (PoE). M ­ VAE7, which is also based on PoE, considers only a partial combination of observed modalities, thereby reducing the number of parameters and improving the computational efficiency. ­Reference8 uses the Mixture-of-Experts (MoE) approach to learn . . .

 



If you want to have access to all the content you need to log in!

Thanks :)

Username or Email

Password

Remember me

Lost your password?

If you don't have an account, you can create one here.