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

Designing neural network architectures is a challenging task and knowingwhich specific layers of a model must be adapted to improve the performance isalmost a mystery. In this paper, we introduce a novel methodology to identifylayers that decrease the test accuracy of trained models. Conflicting layersare detected as early as the beginning of training. In the worst-case scenario,we prove that such a layer could lead to a network that cannot be trained atall. A theoretical analysis is provided on what is the origin of those layersthat result in a lower overall network performance, which is complemented byour extensive empirical evaluation. More precisely, we identified those layersthat worsen the performance because they would produce what we name conflictingtraining bundles. We will show that around 60% of the layers of trainedresidual networks can be completely removed from the architecture with nosignificant increase in the test-error. We will further present a novelneural-architecture-search (NAS) algorithm that identifies conflicting layersat the beginning of the training. Architectures found by our auto-tuningalgorithm achieve competitive accuracy values when compared against morecomplex state-of-the-art architectures, while drastically reducing memoryconsumption and inference time for different computer vision tasks. The sourcecode is available on https://github.com/peerdavid/conflicting-bundles

Auto-tuning of Deep Neural Networks by Conflicting Layer Removal