From e1996dd103f5aab09ded5ad7d001fa6572220164 Mon Sep 17 00:00:00 2001
From: Rain&Bus <rainandbus@gmail.com>
Date: Mon, 1 Apr 2024 18:46:15 +0800
Subject: [PATCH] vault backup: 2024-04-01 18:46:15

---
 .obsidian/appearance.json        | 2 +-
 Books/动手学深度学习/基础概念.md | 2 ++
 2 files changed, 3 insertions(+), 1 deletion(-)

diff --git a/.obsidian/appearance.json b/.obsidian/appearance.json
index e39d435..6ea92b8 100644
--- a/.obsidian/appearance.json
+++ b/.obsidian/appearance.json
@@ -2,7 +2,7 @@
   "accentColor": "",
   "cssTheme": "Minimal",
   "monospaceFontFamily": "Maple Mono SC NF",
-  "theme": "moonstone",
+  "theme": "system",
   "interfaceFontFamily": "霞鹜文楷",
   "textFontFamily": "霞鹜文楷等宽",
   "translucency": false
diff --git a/Books/动手学深度学习/基础概念.md b/Books/动手学深度学习/基础概念.md
index 5b97f76..87573e9 100644
--- a/Books/动手学深度学习/基础概念.md
+++ b/Books/动手学深度学习/基础概念.md
@@ -23,4 +23,6 @@ $L(\mathbf{w}, b) = \frac1n\sum_{i=1}^{n} l^i(\mathbf{x}, b)$
 	反向传播：$w_i ← w_i - \alpha(\frac{\partial L_0}{\partial w_i} + \lambda w_i)$
 	通过反向传播的过程我们可以看到，每次迭代后，权重的值都会变为$(1 - \lambda \alpha)$倍，使得权重值更加靠近零，但是不为0，使模型偏向于学习更加简单的、泛化性能更高的模型。L1正则化则会导致模型将权重集中在一部分特征上，将其它权重清零，这称之为特征选择。
 - Dropout
+	*经典泛化理论认为，为了缩小训练和测试性能之间的差距，应该以简单的模型为目标。简单性的另一个角度是平滑性，即函数不应该对其输入的微小变化敏感。*
+	Dropout在计算每一层时
 	
\ No newline at end of file