Rain-Bus/Obsidian
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

vault backup: 2023-10-12 20:56:34

Browse Source
This commit is contained in:
Rain&Bus
2023-10-12 20:56:34 +08:00
parent a2841425e1
commit bde7056b72
2 changed files with 3 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
.obsidian/workspace.json vendored
View File

@@ -77,7 +77,8 @@
} }
], ],
"direction": "horizontal", "direction": "horizontal",
"width": 315.5 "width": 315.5,
"collapsed": true
}, },
"right": { "right": {
"id": "c501495747cfa761", "id": "c501495747cfa761",

1
Books/动手学深度学习/基础概念.md
View File

@@ -10,3 +10,4 @@ $L(\mathbf{w}, b) = \frac1n\sum_{i=1}^{n} l^i(\mathbf{x}, b)$
梯度下降法主要计算损失函数关于模型参数的导数。但是每次计算时候遍历整个数据集，效率会很低。所以每次计算先抽取一个小批量$B$(由固定数量的样本组成)的梯度,然后我们将梯度乘以一个预先确定的正数$\eta$，并从当前采纳数的值中减掉。 梯度下降法主要计算损失函数关于模型参数的导数。但是每次计算时候遍历整个数据集，效率会很低。所以每次计算先抽取一个小批量$B$(由固定数量的样本组成)的梯度,然后我们将梯度乘以一个预先确定的正数$\eta$，并从当前采纳数的值中减掉。
$(\mathbf{w}, b) <- (\mathbf{w},b) - \frac{\eta}{|B|} \sum_{i\in{B}}\partial_{(\mathbf{w}, b)}l^i(\mathbf{w},b)$ $(\mathbf{w}, b) <- (\mathbf{w},b) - \frac{\eta}{|B|} \sum_{i\in{B}}\partial_{(\mathbf{w}, b)}l^i(\mathbf{w},b)$
其中$\eta$代表学习率 其中$\eta$代表学习率
# 激活函数