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

Self-supervised disentangled representation learning is a critical task insequence modeling. The learnt representations contribute to better modelinterpretability as well as the data generation, and improve the sampleefficiency for downstream tasks. We propose a novel sequence representationlearning method, named Contrastively Disentangled Sequential VariationalAutoencoder (C-DSVAE), to extract and separate the static (time-invariant) anddynamic (time-variant) factors in the latent space. Different from previoussequential variational autoencoder methods, we use a novel evidence lower boundwhich maximizes the mutual information between the input and the latentfactors, while penalizes the mutual information between the static and dynamicfactors. We leverage contrastive estimations of the mutual information terms intraining, together with simple yet effective augmentation techniques, tointroduce additional inductive biases. Our experiments show that C-DSVAEsignificantly outperforms the previous state-of-the-art methods on multiplemetrics.

Contrastively Disentangled Sequential Variational Autoencoder