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 Neural Networks (DNNs) are applied in a wide range of usecases. There isan increased demand for deploying DNNs on devices that do not have abundantresources such as memory and computation units. Recently, network compressionthrough a variety of techniques such as pruning and quantization have beenproposed to reduce the resource requirement. A key parameter that all existingcompression techniques are sensitive to is the compression ratio (e.g., pruningsparsity, quantization bitwidth) of each layer. Traditional solutions treat thecompression ratios of each layer as hyper-parameters, and tune them using humanheuristic. Recent researchers start using black-box hyper-parameteroptimizations, but they will introduce new hyper-parameters and have efficiencyissue. In this paper, we propose a framework to jointly prune and quantize theDNNs automatically according to a target model size without using anyhyper-parameters to manually set the compression ratio for each layer. In theexperiments, we show that our framework can compress the weights data ofResNet-50 to be 836$\times$ smaller without accuracy loss on CIFAR-10, andcompress AlexNet to be 205$\times$ smaller without accuracy loss on ImageNetclassification.

Automatic Neural Network Compression by Sparsity-Quantization Joint Learning: A Constrained Optimization-based Approach