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

Pruning - that is, setting a significant subset of the parameters of a neuralnetwork to zero - is one of the most popular methods of model compression. Yet,several recent works have raised the issue that pruning may induce orexacerbate bias in the output of the compressed model. Despite existingevidence for this phenomenon, the relationship between neural network pruningand induced bias is not well-understood. In this work, we systematicallyinvestigate and characterize this phenomenon in Convolutional Neural Networksfor computer vision. First, we show that it is in fact possible to obtainhighly-sparse models, e.g. with less than 10% remaining weights, which do notdecrease in accuracy nor substantially increase in bias when compared to densemodels. At the same time, we also find that, at higher sparsities, prunedmodels exhibit higher uncertainty in their outputs, as well as increasedcorrelations, which we directly link to increased bias. We propose easy-to-usecriteria which, based only on the uncompressed model, establish whether biaswill increase with pruning, and identify the samples most susceptible to biasedpredictions post-compression.

Bias in Pruned Vision Models: In-Depth Analysis and Countermeasures