FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W3009327594> ?p ?o ?g. }

• W3009327594 abstract "Convolutional neural networks (CNNs) achieve state-of-the-art performance at the cost of becoming deeper and larger. Although quantization (both fixed-point and floating-point) has proven effective for reducing storage and memory access, two challenges -- 1) accuracy loss caused by quantization without calibration, fine-tuning or re-training for deep CNNs and 2) hardware inefficiency caused by floating-point quantization -- prevent processors from completely leveraging the benefits. In this paper, we propose a low-precision floating-point quantization oriented processor, named Phoenix, to address the above challenges. We primarily have three key observations: 1) 8-bit floating-point quantization incurs less error than 8-bit fixed-point quantization; 2) without using any calibration, fine-tuning or re-training techniques, normalization before quantization further reduces accuracy degradation; 3) 8-bit floating-point multiplier achieves higher hardware efficiency than 8-bit fixed-point multiplier if the full-precision product is applied. Based on these key observations, we propose a normalization-oriented 8-bit floating-point quantization method to reduce storage and memory access with negligible accuracy loss (within 0.5%/0.3% for top-1/top-5 accuracy, respectively). We further design a hardware processor to address the hardware inefficiency caused by floating-point multiplier. Compared with a state-of-the-art accelerator, Phoenix is 3.32x and 7.45x better in performance with the same core area for AlexNet and VGG16, respectively." @default.
• W3009327594 created "2020-03-13" @default.
• W3009327594 creator A5011122618 @default.
• W3009327594 creator A5020734784 @default.
• W3009327594 creator A5030691366 @default.
• W3009327594 creator A5030749730 @default.
• W3009327594 creator A5052631227 @default.
• W3009327594 creator A5061724331 @default.
• W3009327594 date "2020-02-29" @default.
• W3009327594 modified "2023-09-24" @default.
• W3009327594 title "Phoenix: A Low-Precision Floating-Point Quantization Oriented Architecture for Convolutional Neural Networks" @default.
• W3009327594 cites W1591127930 @default.
• W3009327594 cites W1594170634 @default.
• W3009327594 cites W1686810756 @default.
• W3009327594 cites W1825672851 @default.
• W3009327594 cites W1976011531 @default.
• W3009327594 cites W1998917233 @default.
• W3009327594 cites W2048266589 @default.
• W3009327594 cites W2067523571 @default.
• W3009327594 cites W2094756095 @default.
• W3009327594 cites W2095705004 @default.
• W3009327594 cites W2097117768 @default.
• W3009327594 cites W2109808436 @default.
• W3009327594 cites W2117539524 @default.
• W3009327594 cites W2152839228 @default.
• W3009327594 cites W2163605009 @default.
• W3009327594 cites W2168231600 @default.
• W3009327594 cites W2183341477 @default.
• W3009327594 cites W2193413348 @default.
• W3009327594 cites W2194775991 @default.
• W3009327594 cites W2198190323 @default.
• W3009327594 cites W2207050309 @default.
• W3009327594 cites W2285660444 @default.
• W3009327594 cites W2286365479 @default.
• W3009327594 cites W2289252105 @default.
• W3009327594 cites W2300242332 @default.
• W3009327594 cites W2319920447 @default.
• W3009327594 cites W2442974303 @default.
• W3009327594 cites W2515287984 @default.
• W3009327594 cites W2520083297 @default.
• W3009327594 cites W2541839172 @default.
• W3009327594 cites W2560017826 @default.
• W3009327594 cites W2563587242 @default.
• W3009327594 cites W2565851976 @default.
• W3009327594 cites W2586654419 @default.
• W3009327594 cites W2595614461 @default.
• W3009327594 cites W2606722458 @default.
• W3009327594 cites W2625954420 @default.
• W3009327594 cites W2748818695 @default.
• W3009327594 cites W2762597430 @default.
• W3009327594 cites W2789305180 @default.
• W3009327594 cites W2803935332 @default.
• W3009327594 cites W2883920103 @default.
• W3009327594 cites W2883929540 @default.
• W3009327594 cites W2889797931 @default.
• W3009327594 cites W2899818272 @default.
• W3009327594 cites W2903735800 @default.
• W3009327594 cites W2904902077 @default.
• W3009327594 cites W2946355854 @default.
• W3009327594 cites W2953212265 @default.
• W3009327594 cites W2962761403 @default.
• W3009327594 cites W2963037989 @default.
• W3009327594 cites W2963114950 @default.
• W3009327594 cites W2963122961 @default.
• W3009327594 cites W2963367920 @default.
• W3009327594 cites W2963409685 @default.
• W3009327594 cites W2963446712 @default.
• W3009327594 cites W2963674932 @default.
• W3009327594 cites W2964299589 @default.
• W3009327594 cites W2964350391 @default.
• W3009327594 hasPublicationYear "2020" @default.
• W3009327594 type Work @default.
• W3009327594 sameAs 3009327594 @default.
• W3009327594 citedByCount "0" @default.
• W3009327594 crossrefType "posted-content" @default.
• W3009327594 hasAuthorship W3009327594A5011122618 @default.
• W3009327594 hasAuthorship W3009327594A5020734784 @default.
• W3009327594 hasAuthorship W3009327594A5030691366 @default.
• W3009327594 hasAuthorship W3009327594A5030749730 @default.
• W3009327594 hasAuthorship W3009327594A5052631227 @default.
• W3009327594 hasAuthorship W3009327594A5061724331 @default.
• W3009327594 hasConcept C113775141 @default.
• W3009327594 hasConcept C11413529 @default.
• W3009327594 hasConcept C124584101 @default.
• W3009327594 hasConcept C139719470 @default.
• W3009327594 hasConcept C154945302 @default.
• W3009327594 hasConcept C162324750 @default.
• W3009327594 hasConcept C173608175 @default.
• W3009327594 hasConcept C28855332 @default.
• W3009327594 hasConcept C41008148 @default.
• W3009327594 hasConcept C81363708 @default.
• W3009327594 hasConcept C84211073 @default.
• W3009327594 hasConcept C9390403 @default.
• W3009327594 hasConceptScore W3009327594C113775141 @default.
• W3009327594 hasConceptScore W3009327594C11413529 @default.
• W3009327594 hasConceptScore W3009327594C124584101 @default.
• W3009327594 hasConceptScore W3009327594C139719470 @default.
• W3009327594 hasConceptScore W3009327594C154945302 @default.
• W3009327594 hasConceptScore W3009327594C162324750 @default.
• W3009327594 hasConceptScore W3009327594C173608175 @default.

• next