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) offer the highest performance in a wide range ofapplications in computer vision. These results rely on over-parameterizedbackbones, which are expensive to run. This computational burden can bedramatically reduced by quantizing (in either data-free (DFQ), post-training(PTQ) or quantization-aware training (QAT) scenarios) floating point values toternary values (2 bits, with each weight taking value in {-1,0,1}). In thiscontext, we observe that rounding to nearest minimizes the expected error givena uniform distribution and thus does not account for the skewness and kurtosisof the weight distribution, which strongly affects ternary quantizationperformance. This raises the following question: shall one minimize the highestor average quantization error? To answer this, we design two operators: TQuantand MQuant that correspond to these respective minimization tasks. We showexperimentally that our approach allows to significantly improve theperformance of ternary quantization through a variety of scenarios in DFQ, PTQand QAT and give strong insights to pave the way for future research in deepneural network quantization.

Designing strong baselines for ternary neural network quantization through support and mass equalization