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

The demand for efficient processing of deep neural networks (DNNs) onembedded devices is a significant challenge limiting their deployment.Exploiting sparsity in the network's feature maps is one of the ways to reduceits inference latency. It is known that unstructured sparsity results in loweraccuracy degradation with respect to structured sparsity but the former needsextensive inference engine changes to get latency benefits. To tackle thischallenge, we propose a solution to induce semi-structured activation sparsityexploitable through minor runtime modifications. To attain high speedup levelsat inference time, we design a sparse training procedure with awareness of thefinal position of the activations while computing the General MatrixMultiplication (GEMM). We extensively evaluate the proposed solution acrossvarious models for image classification and object detection tasks. Remarkably,our approach yields a speed improvement of $1.25 \times$ with a minimalaccuracy drop of $1.1\%$ for the ResNet18 model on the ImageNet dataset.Furthermore, when combined with a state-of-the-art structured pruning method,the resulting models provide a good latency-accuracy trade-off, outperformingmodels that solely employ structured pruning techniques.

Accelerating Deep Neural Networks via Semi-Structured Activation Sparsity