We propose a new model for relational VAE semi-supervision capable of balancing disentanglement and low complexity modelling of relations with different symbolic properties. We compare the relative benefits of relation-decoder complexity and latent space structure on both inductive and transductive transfer learning. Our results depict a complex picture where enforcing structure on semi-supervised representations can greatly improve zero-shot transductive transfer, but may be less favourable or even impact negatively the capacity for inductive transfer.
@article{arxiv.2011.07137,
title = {On the Transferability of VAE Embeddings using Relational Knowledge with Semi-Supervision},
author = {Harald Strömfelt and Luke Dickens and Artur d'Avila Garcez and Alessandra Russo},
journal= {arXiv preprint arXiv:2011.07137},
year = {2020}
}