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

The conventional wisdom behind learning deep classification models is tofocus on bad-classified examples and ignore well-classified examples that arefar from the decision boundary. For instance, when training with cross-entropyloss, examples with higher likelihoods (i.e., well-classified examples)contribute smaller gradients in back-propagation. However, we theoreticallyshow that this common practice hinders representation learning, energyoptimization, and margin growth. To counteract this deficiency, we propose toreward well-classified examples with additive bonuses to revive theircontribution to the learning process. This counterexample theoreticallyaddresses these three issues. We empirically support this claim by directlyverifying the theoretical results or significant performance improvement withour counterexample on diverse tasks, including image classification, graphclassification, and machine translation. Furthermore, this paper shows that wecan deal with complex scenarios, such as imbalanced classification, OODdetection, and applications under adversarial attacks because our idea cansolve these three issues. Code is available at:https://github.com/lancopku/well-classified-examples-are-underestimated.

Well-classified Examples are Underestimated in Classification with Deep Neural Networks