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

Convolutional neural networks (CNNs) achieve remarkable performance in a widerange of fields. However, intensive memory access of activations introducesconsiderable energy consumption, impeding deployment of CNNs onresourceconstrained edge devices. Existing works in activation compressionpropose to transform feature maps for higher compressibility, thus enablingdimension reduction. Nevertheless, in the case of aggressive dimensionreduction, these methods lead to severe accuracy drop. To improve the trade-offbetween classification accuracy and compression ratio, we propose acompression-aware projection system, which employs a learnable projection tocompensate for the reconstruction loss. In addition, a greedy selection metricis introduced to optimize the layer-wise compression ratio allocation byconsidering both accuracy and #bits reduction simultaneously. Our test resultsshow that the proposed methods effectively reduce 2.91x~5.97x memory accesswith negligible accuracy drop on MobileNetV2/ResNet18/VGG16.

Compression-aware Projection with Greedy Dimension Reduction for Convolutional Neural Network Activations