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 Tools annual

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 Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Overfitting is one of the most common problems when training deep neuralnetworks on comparatively small datasets. Here, we demonstrate that neuralnetwork activation sparsity is a reliable indicator for overfitting which weutilize to propose novel targeted sparsity visualization and regularizationstrategies. Based on these strategies we are able to understand and counteractoverfitting caused by activation sparsity and filter correlation in a targetedlayer-by-layer manner. Our results demonstrate that targeted sparsityregularization can efficiently be used to regularize well-known datasets andarchitectures with a significant increase in image classification performancewhile outperforming both dropout and batch normalization. Ultimately, our studyreveals novel insights into the contradicting concepts of activation sparsityand network capacity by demonstrating that targeted sparsity regularizationenables salient and discriminative feature learning while exploiting the fullcapacity of deep models without suffering from overfitting, even when trainedexcessively.

Exploiting the Full Capacity of Deep Neural Networks while Avoiding Overfitting by Targeted Sparsity Regularization