Matches in SemOpenAlex for { <https://semopenalex.org/work/W4285050159> ?p ?o ?g. }
Showing items 1 to 58 of
58
with 100 items per page.
- W4285050159 abstract "The compression of deep learning models is of fundamental importance in deploying such models to edge devices. The selection of compression parameters can be automated to meet changes in the hardware platform and application using optimization algorithms. This article introduces a Multi-Objective Hardware-Aware Quantization (MOHAQ) method, which considers hardware efficiency and inference error as objectives for mixed-precision quantization. The proposed method feasibly evaluates candidate solutions in a large search space by relying on two steps. First, post-training quantization is applied for fast solution evaluation (inference-only search). Second, we propose the beacon-based search to retrain selected solutions only and use them as beacons to know the effect of retraining on other solutions. We use a speech recognition model based on Simple Recurrent Unit (SRU) using the TIMIT dataset and apply our method to run on SiLago and Bitfusion platforms. We provide experimental evaluations showing that SRU can be compressed up to 8x by post-training quantization without any significant error increase. On SiLago, we found solutions that achieve 97% and 86% of the maximum possible speedup and energy saving, with a minor increase in error. On Bitfusion, beacon-based search reduced the error gain of inference-only search by up to 4.9 percentage points." @default.
- W4285050159 created "2022-07-13" @default.
- W4285050159 creator A5007772917 @default.
- W4285050159 creator A5020628204 @default.
- W4285050159 creator A5021872962 @default.
- W4285050159 creator A5026355063 @default.
- W4285050159 creator A5044683684 @default.
- W4285050159 creator A5048826742 @default.
- W4285050159 date "2021-08-02" @default.
- W4285050159 modified "2023-10-15" @default.
- W4285050159 title "MOHAQ: Multi-Objective Hardware-Aware Quantization of Recurrent Neural Networks" @default.
- W4285050159 doi "https://doi.org/10.48550/arxiv.2108.01192" @default.
- W4285050159 hasPublicationYear "2021" @default.
- W4285050159 type Work @default.
- W4285050159 citedByCount "0" @default.
- W4285050159 crossrefType "posted-content" @default.
- W4285050159 hasAuthorship W4285050159A5007772917 @default.
- W4285050159 hasAuthorship W4285050159A5020628204 @default.
- W4285050159 hasAuthorship W4285050159A5021872962 @default.
- W4285050159 hasAuthorship W4285050159A5026355063 @default.
- W4285050159 hasAuthorship W4285050159A5044683684 @default.
- W4285050159 hasAuthorship W4285050159A5048826742 @default.
- W4285050159 hasBestOaLocation W42850501591 @default.
- W4285050159 hasConcept C113775141 @default.
- W4285050159 hasConcept C11413529 @default.
- W4285050159 hasConcept C119857082 @default.
- W4285050159 hasConcept C154945302 @default.
- W4285050159 hasConcept C173608175 @default.
- W4285050159 hasConcept C2776214188 @default.
- W4285050159 hasConcept C28855332 @default.
- W4285050159 hasConcept C41008148 @default.
- W4285050159 hasConcept C68339613 @default.
- W4285050159 hasConceptScore W4285050159C113775141 @default.
- W4285050159 hasConceptScore W4285050159C11413529 @default.
- W4285050159 hasConceptScore W4285050159C119857082 @default.
- W4285050159 hasConceptScore W4285050159C154945302 @default.
- W4285050159 hasConceptScore W4285050159C173608175 @default.
- W4285050159 hasConceptScore W4285050159C2776214188 @default.
- W4285050159 hasConceptScore W4285050159C28855332 @default.
- W4285050159 hasConceptScore W4285050159C41008148 @default.
- W4285050159 hasConceptScore W4285050159C68339613 @default.
- W4285050159 hasLocation W42850501591 @default.
- W4285050159 hasLocation W42850501592 @default.
- W4285050159 hasOpenAccess W4285050159 @default.
- W4285050159 hasPrimaryLocation W42850501591 @default.
- W4285050159 hasRelatedWork W2752721426 @default.
- W4285050159 hasRelatedWork W3008756425 @default.
- W4285050159 hasRelatedWork W3184547836 @default.
- W4285050159 hasRelatedWork W3195170785 @default.
- W4285050159 hasRelatedWork W3195381335 @default.
- W4285050159 hasRelatedWork W3196579076 @default.
- W4285050159 hasRelatedWork W4224917404 @default.
- W4285050159 hasRelatedWork W4286233754 @default.
- W4285050159 hasRelatedWork W4287864142 @default.
- W4285050159 hasRelatedWork W4307933444 @default.
- W4285050159 isParatext "false" @default.
- W4285050159 isRetracted "false" @default.
- W4285050159 workType "article" @default.