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

Deep neural networks tend to make overconfident predictions and often requireadditional detectors for misclassifications, particularly for safety-criticalapplications. Existing detection methods usually only focus on adversarialattacks or out-of-distribution samples as reasons for false predictions.However, generalization errors occur due to diverse reasons often related topoorly learning relevant invariances. We therefore propose GIT, a holisticapproach for the detection of generalization errors that combines the usage ofgradient information and invariance transformations. The invariancetransformations are designed to shift misclassified samples back into thegeneralization area of the neural network, while the gradient informationmeasures the contradiction between the initial prediction and the correspondinginherent computations of the neural network using the transformed sample. Ourexperiments demonstrate the superior performance of GIT compared to thestate-of-the-art on a variety of network architectures, problem setups andperturbation types.

GIT: Detecting Uncertainty, Out-Of-Distribution and Adversarial Samples using Gradients and Invariance Transformations