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

Ensembling a neural network is a widely recognized approach to enhance modelperformance, estimate uncertainty, and improve robustness in deep supervisedlearning. However, deep ensembles often come with high computational costs andmemory demands. In addition, the efficiency of a deep ensemble is related todiversity among the ensemble members which is challenging for large,over-parameterized deep neural networks. Moreover, ensemble learning has notyet seen such widespread adoption, and it remains a challenging endeavor forself-supervised or unsupervised representation learning. Motivated by thesechallenges, we present a novel self-supervised training regime that leveragesan ensemble of independent sub-networks, complemented by a new loss functiondesigned to encourage diversity. Our method efficiently builds a sub-modelensemble with high diversity, leading to well-calibrated estimates of modeluncertainty, all achieved with minimal computational overhead compared totraditional deep self-supervised ensembles. To evaluate the effectiveness ofour approach, we conducted extensive experiments across various tasks,including in-distribution generalization, out-of-distribution detection,dataset corruption, and semi-supervised settings. The results demonstrate thatour method significantly improves prediction reliability. Our approach not onlyachieves excellent accuracy but also enhances calibration, surpassing baselineperformance across a wide range of self-supervised architectures in computervision, natural language processing, and genomics data.

Diversified Ensemble of Independent Sub-Networks for Robust Self-Supervised Representation Learning