30
/ru/
AIzaSyAYiBZKx7MnpbEhh9jyipgxe19OcubqV5w
April 1, 2024
3251714
288434
1
Public Timelines FAQ
Public TimelinesFAQ
Public TimelinesFAQ
For educational institutions For teachers For students Open cabinet
For educational institutions For teachers For students Open cabinet
Создать
Create a timeline
Public timelines
Library
FAQ

5 янв 2017 г. - Deep Learning with Low Precision by Half-wave Gaussian Quantization (HWGQ)

Описание:

本文主要讨论了在低位宽量化的近似梯度训练中，梯度消失的问题，通过设计新的量化函数(sign/Q)和激活函数来保证梯度不消失，以实现量化模型的收敛。有趣的是，仅通过对relu函数的剪切或者log-tail变换，模型的性能得到了显著提升。这或许就是特征的影响

Добавлено на ленту времени:

Quantization timeline

By龚成
29 мар 2020

Дата:

5 янв 2017 г.
Сейчас
~ 7 гг и 4 мес назад
About & FeedbackСоглашениеПриватностьБиблиотека
2024 © Time.Graphics
Support 24/7  
CabinetGet premium
Donate
The service accepts bank transfer (ACH, Wire) or cards (Visa, MasterCard, etc). Processed by Stripe.
Secured with SSL
Excellent (Trustpilot Reviews)
Based on 115+ reviews
Write your own review on Trustpilot.com