Matches in SemOpenAlex for { <https://semopenalex.org/work/W4320855500> ?p ?o ?g. }
Showing items 1 to 59 of
59
with 100 items per page.
- W4320855500 abstract "Quantized neural networks are well known for reducing latency, power consumption, and model size without significant degradation in accuracy, making them highly applicable for systems with limited resources and low power requirements. Mixed precision quantization offers better utilization of customized hardware that supports arithmetic operations at different bitwidths. Existing mixed-precision schemes rely on having a high exploration space, resulting in a large carbon footprint. In addition, these bit allocation strategies mostly induce constraints on the model size rather than utilizing the performance of neural network deployment on specific hardware. Our work proposes Fast-Bit Allocation for Mixed-Precision Quantization (FBM), which finds an optimal bitwidth allocation by measuring desired behaviors through a simulation of a specific device, or even on a physical one. While dynamic transitions of bit allocation in mixed precision quantization with ultra-low bitwidth are known to suffer from performance degradation, we present a fast recovery solution from such transitions. A comprehensive evaluation of the proposed method on CIFAR-10 and ImageNet demonstrates our method's superiority over current state-of-the-art schemes in terms of the trade-off between neural network accuracy and hardware efficiency. Our source code, experimental settings and quantized models are available at https://github.com/RamorayDrake/FBM/" @default.
- W4320855500 created "2023-02-16" @default.
- W4320855500 creator A5001688990 @default.
- W4320855500 creator A5019913171 @default.
- W4320855500 creator A5046832314 @default.
- W4320855500 creator A5063770531 @default.
- W4320855500 creator A5066356247 @default.
- W4320855500 creator A5089135250 @default.
- W4320855500 date "2022-05-30" @default.
- W4320855500 modified "2023-09-27" @default.
- W4320855500 title "FBM: Fast-Bit Allocation for Mixed-Precision Quantization" @default.
- W4320855500 doi "https://doi.org/10.48550/arxiv.2205.15437" @default.
- W4320855500 hasPublicationYear "2022" @default.
- W4320855500 type Work @default.
- W4320855500 citedByCount "0" @default.
- W4320855500 crossrefType "posted-content" @default.
- W4320855500 hasAuthorship W4320855500A5001688990 @default.
- W4320855500 hasAuthorship W4320855500A5019913171 @default.
- W4320855500 hasAuthorship W4320855500A5046832314 @default.
- W4320855500 hasAuthorship W4320855500A5063770531 @default.
- W4320855500 hasAuthorship W4320855500A5066356247 @default.
- W4320855500 hasAuthorship W4320855500A5089135250 @default.
- W4320855500 hasBestOaLocation W43208555001 @default.
- W4320855500 hasConcept C105339364 @default.
- W4320855500 hasConcept C111919701 @default.
- W4320855500 hasConcept C113775141 @default.
- W4320855500 hasConcept C11413529 @default.
- W4320855500 hasConcept C154945302 @default.
- W4320855500 hasConcept C28855332 @default.
- W4320855500 hasConcept C41008148 @default.
- W4320855500 hasConcept C50644808 @default.
- W4320855500 hasConcept C76155785 @default.
- W4320855500 hasConcept C82876162 @default.
- W4320855500 hasConceptScore W4320855500C105339364 @default.
- W4320855500 hasConceptScore W4320855500C111919701 @default.
- W4320855500 hasConceptScore W4320855500C113775141 @default.
- W4320855500 hasConceptScore W4320855500C11413529 @default.
- W4320855500 hasConceptScore W4320855500C154945302 @default.
- W4320855500 hasConceptScore W4320855500C28855332 @default.
- W4320855500 hasConceptScore W4320855500C41008148 @default.
- W4320855500 hasConceptScore W4320855500C50644808 @default.
- W4320855500 hasConceptScore W4320855500C76155785 @default.
- W4320855500 hasConceptScore W4320855500C82876162 @default.
- W4320855500 hasLocation W43208555001 @default.
- W4320855500 hasOpenAccess W4320855500 @default.
- W4320855500 hasPrimaryLocation W43208555001 @default.
- W4320855500 hasRelatedWork W1987753576 @default.
- W4320855500 hasRelatedWork W2362198170 @default.
- W4320855500 hasRelatedWork W2384390289 @default.
- W4320855500 hasRelatedWork W2386387936 @default.
- W4320855500 hasRelatedWork W3036033155 @default.
- W4320855500 hasRelatedWork W3185695382 @default.
- W4320855500 hasRelatedWork W4294982680 @default.
- W4320855500 hasRelatedWork W4311730469 @default.
- W4320855500 hasRelatedWork W4312263439 @default.
- W4320855500 hasRelatedWork W2779562428 @default.
- W4320855500 isParatext "false" @default.
- W4320855500 isRetracted "false" @default.
- W4320855500 workType "article" @default.