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

Optimization in the latent space of variational autoencoders is a promisingapproach to generate high-dimensional discrete objects that maximize anexpensive black-box property (e.g., drug-likeness in molecular generation,function approximation with arithmetic expressions). However, existing methodslack robustness as they may decide to explore areas of the latent space forwhich no data was available during training and where the decoder can beunreliable, leading to the generation of unrealistic or invalid objects. Wepropose to leverage the epistemic uncertainty of the decoder to guide theoptimization process. This is not trivial though, as a naive estimation ofuncertainty in the high-dimensional and structured settings we consider wouldresult in high estimator variance. To solve this problem, we introduce animportance sampling-based estimator that provides more robust estimates ofepistemic uncertainty. Our uncertainty-guided optimization approach does notrequire modifications of the model architecture nor the training process. Itproduces samples with a better trade-off between black-box objective andvalidity of the generated samples, sometimes improving both simultaneously. Weillustrate these advantages across several experimental settings in digitgeneration, arithmetic expression approximation and molecule generation fordrug design.

Improving black-box optimization in VAE latent space using decoder uncertainty