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

Quantization and pruning are two effective Deep Neural Networks modelcompression methods. In this paper, we propose Automatic Prune Binarization(APB), a novel compression technique combining quantization with pruning. APBenhances the representational capability of binary networks using a fewfull-precision weights. Our technique jointly maximizes the accuracy of thenetwork while minimizing its memory impact by deciding whether each weightshould be binarized or kept in full precision. We show how to efficientlyperform a forward pass through layers compressed using APB by decomposing itinto a binary and a sparse-dense matrix multiplication. Moreover, we design twonovel efficient algorithms for extremely quantized matrix multiplication onCPU, leveraging highly efficient bitwise operations. The proposed algorithmsare 6.9x and 1.5x faster than available state-of-the-art solutions. Weextensively evaluate APB on two widely adopted model compression datasets,namely CIFAR10 and ImageNet. APB delivers better accuracy/memory trade-offcompared to state-of-the-art methods based on i) quantization, ii) pruning, andiii) combination of pruning and quantization. APB outperforms quantization inthe accuracy/efficiency trade-off, being up to 2x faster than the 2-bitquantized model with no loss in accuracy.

Neural Network Compression using Binarization and Few Full-Precision Weights