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 compression of deep learning models is of fundamental importance indeploying such models to edge devices. The selection of compression parameterscan be automated to meet changes in the hardware platform and application usingoptimization algorithms. This article introduces a Multi-ObjectiveHardware-Aware Quantization (MOHAQ) method, which considers hardware efficiencyand inference error as objectives for mixed-precision quantization. Theproposed method feasibly evaluates candidate solutions in a large search spaceby relying on two steps. First, post-training quantization is applied for fastsolution evaluation (inference-only search). Second, we propose the"beacon-based search" to retrain selected solutions only and use them asbeacons to know the effect of retraining on other solutions. We use a speechrecognition model based on Simple Recurrent Unit (SRU) using the TIMIT datasetand apply our method to run on SiLago and Bitfusion platforms. We provideexperimental evaluations showing that SRU can be compressed up to 8x bypost-training quantization without any significant error increase. On SiLago,we found solutions that achieve 97\% and 86\% of the maximum possible speedupand energy saving, with a minor increase in error. On Bitfusion, beacon-basedsearch reduced the error gain of inference-only search by up to 4.9 percentagepoints.

MOHAQ: Multi-Objective Hardware-Aware Quantization of Recurrent Neural Networks