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 Tools annual

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 Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

While generative models have shown great success in generatinghigh-dimensional samples conditional on low-dimensional descriptors (learninge.g. stroke thickness in MNIST, hair color in CelebA, or speaker identity inWavenet), their generation out-of-sample poses fundamental problems. Theconditional variational autoencoder (CVAE) as a simple conditional generativemodel does not explicitly relate conditions during training and, hence, has noincentive of learning a compact joint distribution across conditions. Weovercome this limitation by matching their distributions using maximum meandiscrepancy (MMD) in the decoder layer that follows the bottleneck. Thisintroduces a strong regularization both for reconstructing samples within thesame condition and for transforming samples across conditions, resulting inmuch improved generalization. We refer to the architecture as\emph{transformer} VAE (trVAE). Benchmarking trVAE on high-dimensional imageand tabular data, we demonstrate higher robustness and higher accuracy thanexisting approaches. In particular, we show qualitatively improved predictionsfor cellular perturbation response to treatment and disease based onhigh-dimensional single-cell gene expression data, by tackling previouslyproblematic minority classes and multiple conditions. For generic tasks, weimprove Pearson correlations of high-dimensional estimated means and varianceswith their ground truths from 0.89 to 0.97 and 0.75 to 0.87, respectively.

Conditional out-of-sample generation for unpaired data using trVAE