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 learning approaches have provided state-of-the-art performance in manyapplications by relying on large and overparameterized neural networks.However, such networks have been shown to be very brittle and are difficult todeploy on resource-limited platforms. Model pruning, i.e., reducing the size ofthe network, is a widely adopted strategy that can lead to a more robust andcompact model. Many heuristics exist for model pruning, but empirical studiesshow that some heuristics improve performance whereas others can make modelsmore brittle or have other side effects. This work aims to shed light on howdifferent pruning methods alter the network's internal feature representationand the corresponding impact on model performance. To facilitate acomprehensive comparison and characterization of the high-dimensional modelfeature space, we introduce a visual geometric analysis of featurerepresentations. We decomposed and evaluated a set of critical geometricconcepts from the common adopted classification loss, and used them to design avisualization system to compare and highlight the impact of pruning on modelperformance and feature representation. The proposed tool provides anenvironment for in-depth comparison of pruning methods and a comprehensiveunderstanding of how model response to common data corruption. By leveragingthe proposed visualization, machine learning researchers can reveal thesimilarities between pruning methods and redundant in robustness evaluationbenchmarks, obtain geometric insights about the differences between prunedmodels that achieve superior robustness performance, and identify samples thatare robust or fragile to model pruning and common data corruption to modelpruning and data corruption but also obtain insights and explanations on howsome pruned models achieve superior robustness performance.

"Understanding Robustness Lottery": A Geometric Visual Comparative Analysis of Neural Network Pruning Approaches