Skip to content

Commit

Permalink
结构调整
Browse files Browse the repository at this point in the history
  • Loading branch information
imhuay committed Sep 24, 2018
1 parent f5e8fd1 commit 82274e2
Show file tree
Hide file tree
Showing 72 changed files with 19 additions and 8 deletions.
15 changes: 7 additions & 8 deletions A-机器学习/A-机器学习基础.md
Original file line number Diff line number Diff line change
Expand Up @@ -6,11 +6,11 @@ Index
<!-- TOC -->

- [偏差与方差](#偏差与方差)
- [导致偏差和方差的原因](#导致偏差和方差的原因)
- [深度学习中的偏差与方差](#深度学习中的偏差与方差)
- [偏差/方差 与 Boosting/Bagging](#偏差方差-与-boostingbagging)
- [偏差与方差的计算公式](#偏差与方差的计算公式)
- [偏差与方差的权衡(过拟合与模型复杂度的权衡)](#偏差与方差的权衡过拟合与模型复杂度的权衡)
- [导致偏差和方差的原因](#导致偏差和方差的原因)
- [深度学习中的偏差与方差](#深度学习中的偏差与方差)
- [偏差/方差 与 Boosting/Bagging](#偏差方差-与-boostingbagging)
- [偏差与方差的计算公式](#偏差与方差的计算公式)
- [偏差与方差的权衡(过拟合与模型复杂度的权衡)](#偏差与方差的权衡过拟合与模型复杂度的权衡)
- [生成模型与判别模型](#生成模型与判别模型)
- [先验概率与后验概率](#先验概率与后验概率)

Expand Down Expand Up @@ -41,10 +41,10 @@ Index
- 神经网络的拟合能力非常强,因此它的**训练误差**(偏差)通常较小;
- 但是过强的拟合能力会导致较大的方差,使模型的测试误差(**泛化误差**)增大;
- 因此深度学习的核心工作之一就是研究如何降低模型的泛化误差,这类方法统称为**正则化方法**
> ./深度学习基础/[正则化](./DL-A-深度学习基础.md#正则化)
> ../深度学习/[正则化](../B-深度学习/C-专题-正则化)
### 偏差/方差 与 Boosting/Bagging
> ./集成学习专题/[Boosting/Bagging 与 偏差/方差 的关系](./ML-B-专题-集成学习.md#boostingbagging-与-偏差方差-的关系)
> ./集成学习专题/[Boosting/Bagging 与 偏差/方差 的关系](./C-专题-集成学习#boostingbagging-与-偏差方差-的关系)
### 偏差与方差的计算公式
- 记在**训练集 D** 上学得的模型为
Expand All @@ -69,7 +69,6 @@ Index
-**偏差-方差分解**”表明模型的泛化能力是由算法的能力、数据的充分性、任务本身的难度共同决定的。

### 偏差与方差的权衡(过拟合与模型复杂度的权衡)
> ./深度学习基础/[过拟合与欠拟合](./DL-A-深度学习基础.md#过拟合与欠拟合)
- 给定学习任务,
- 当训练不足时,模型的**拟合能力不够**(数据的扰动不足以使模型产生显著的变化),此时**偏差**主导模型的泛化误差;
- 随着训练的进行,模型的**拟合能力增强**(模型能够学习数据发生的扰动),此时**方差**逐渐主导模型的泛化误差;
Expand Down
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
12 changes: 12 additions & 0 deletions B-深度学习/C-专题-正则化.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,12 @@
专题-正则化
===

Reference
---
- 知乎/[详解深度学习中的 Normalization](https://zhuanlan.zhihu.com/p/33173246)

Index
---


##
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.

0 comments on commit 82274e2

Please sign in to comment.