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

Contrastive learning relies on constructing a collection of negative examplesthat are sufficiently hard to discriminate against positive queries when theirrepresentations are self-trained. Existing contrastive learning methods eithermaintain a queue of negative samples over minibatches while only a smallportion of them are updated in an iteration, or only use the other examplesfrom the current minibatch as negatives. They could not closely track thechange of the learned representation over iterations by updating the entirequeue as a whole, or discard the useful information from the past minibatches.Alternatively, we present to directly learn a set of negative adversariesplaying against the self-trained representation. Two players, therepresentation network and negative adversaries, are alternately updated toobtain the most challenging negative examples against which the representationof positive queries will be trained to discriminate. We further show that thenegative adversaries are updated towards a weighted combination of positivequeries by maximizing the adversarial contrastive loss, thereby allowing themto closely track the change of representations over time. Experiment resultsdemonstrate the proposed Adversarial Contrastive (AdCo) model not only achievessuperior performances (a top-1 accuracy of 73.2\% over 200 epochs and 75.7\%over 800 epochs with linear evaluation on ImageNet), but also can bepre-trained more efficiently with fewer epochs.

AdCo: Adversarial Contrast for Efficient Learning of Unsupervised Representations from Self-Trained Negative Adversaries