Jump to content
Toggle sidebar
Marovi AI
Search
English
Personal tools
Create account
Log in
Navigation
Main page
Recent changes
Random page
Help about MediaWiki
Tools
What links here
Related changes
Special pages
Printable version
Permanent link
Page information
In other languages
Add languages
Translations
:
Stochastic Gradient Descent/27/zh
Translation unit
Discussion
不转换
不转换
简体
繁體
大陆简体
香港繁體
澳門繁體
大马简体
新加坡简体
臺灣正體
Read
View source
View history
More
Read
View source
View history
From Marovi AI
Revision as of 03:38, 27 April 2026 by
DeployBot
(
talk
|
contribs
)
(Batch translate Stochastic Gradient Descent unit 27 → zh)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
数据洗牌
—— 在每个 epoch 重新打乱数据集,避免出现循环模式。
梯度裁剪
—— 对梯度范数进行截断,以防止更新爆炸,尤其是在循环神经网络中。
批归一化
—— 对层输入进行归一化可降低对
学习率
的敏感度。
混合精度训练
—— 使用半精度浮点数能在现代 GPU 上加速 SGD,同时几乎不损失精度。