English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Disentangled generative models map a latent code vector to a target space,while enforcing that a subset of the learned latent codes are interpretable andassociated with distinct properties of the target distribution. Recent advanceshave been dominated by Variational AutoEncoder (VAE)-based methods, whiletraining disentangled generative adversarial networks (GANs) remainschallenging. In this work, we show that the dominant challenges facingdisentangled GANs can be mitigated through the use of self-supervision. We maketwo main contributions: first, we design a novel approach for trainingdisentangled GANs with self-supervision. We propose contrastive regularizer,which is inspired by a natural notion of disentanglement: latent traversal.This achieves higher disentanglement scores than state-of-the-art VAE- andGAN-based approaches. Second, we propose an unsupervised model selection schemecalled ModelCentrality, which uses generated synthetic samples to compute themedoid (multi-dimensional generalization of median) of a collection of models.The current common practice of hyper-parameter tuning requires usingground-truths samples, each labelled with known perfect disentangled latentcodes. As real datasets are not equipped with such labels, we propose anunsupervised model selection scheme and show that it finds a model close to thebest one, for both VAEs and GANs. Combining contrastive regularization withModelCentrality, we improve upon the state-of-the-art disentanglement scoressignificantly, without accessing the supervised data.

InfoGAN-CR and ModelCentrality: Self-supervised Model Training and Selection for Disentangling GANs