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 methods have shown to be effective at reducing the size of deepneural networks while keeping accuracy almost intact. Among the most effectivemethods are those that prune a network while training it with a sparsity priorloss and learnable dropout parameters. A shortcoming of these approacheshowever is that neither the size nor the inference speed of the pruned networkcan be controlled directly; yet this is a key feature for targeting deploymentof CNNs on low-power hardware. To overcome this, we introduce a budgetedregularized pruning framework for deep CNNs. Our approach naturally fits intotraditional neural network training as it consists of a learnable maskinglayer, a novel budget-aware objective function, and the use of knowledgedistillation. We also provide insights on how to prune a residual network andhow this can lead to new architectures. Experimental results reveal that CNNspruned with our method are more accurate and less compute-hungry thanstate-of-the-art methods. Also, our approach is more effective at preventingaccuracy collapse in case of severe pruning; this allows us to attain pruningfactors up to 16x without significant accuracy drop.

Structured Pruning of Neural Networks with Budget-Aware Regularization