賴冠州
    • Create new note
    • Create a note from template
      • Sharing URL Link copied
      • /edit
      • View mode
        • Edit mode
        • View mode
        • Book mode
        • Slide mode
        Edit mode View mode Book mode Slide mode
      • Customize slides
      • Note Permission
      • Read
        • Only me
        • Signed-in users
        • Everyone
        Only me Signed-in users Everyone
      • Write
        • Only me
        • Signed-in users
        • Everyone
        Only me Signed-in users Everyone
      • Engagement control Commenting, Suggest edit, Emoji Reply
      • Invitee
    • Publish Note

      Share your work with the world Congratulations! 🎉 Your note is out in the world Publish Note

      Your note will be visible on your profile and discoverable by anyone.
      Your note is now live.
      This note is visible on your profile and discoverable online.
      Everyone on the web can find and read all notes of this public team.
      See published notes
      Unpublish note
      Please check the box to agree to the Community Guidelines.
      View profile
    • Commenting
      Permission
      Disabled Forbidden Owners Signed-in users Everyone
    • Enable
    • Permission
      • Forbidden
      • Owners
      • Signed-in users
      • Everyone
    • Suggest edit
      Permission
      Disabled Forbidden Owners Signed-in users Everyone
    • Enable
    • Permission
      • Forbidden
      • Owners
      • Signed-in users
    • Emoji Reply
    • Enable
    • Versions and GitHub Sync
    • Note settings
    • Engagement control
    • Transfer ownership
    • Delete this note
    • Save as template
    • Insert from template
    • Import from
      • Dropbox
      • Google Drive
      • Gist
      • Clipboard
    • Export to
      • Dropbox
      • Google Drive
      • Gist
    • Download
      • Markdown
      • HTML
      • Raw HTML
Menu Note settings Sharing URL Create Help
Create Create new note Create a note from template
Menu
Options
Versions and GitHub Sync Engagement control Transfer ownership Delete this note
Import from
Dropbox Google Drive Gist Clipboard
Export to
Dropbox Google Drive Gist
Download
Markdown HTML Raw HTML
Back
Sharing URL Link copied
/edit
View mode
  • Edit mode
  • View mode
  • Book mode
  • Slide mode
Edit mode View mode Book mode Slide mode
Customize slides
Note Permission
Read
Only me
  • Only me
  • Signed-in users
  • Everyone
Only me Signed-in users Everyone
Write
Only me
  • Only me
  • Signed-in users
  • Everyone
Only me Signed-in users Everyone
Engagement control Commenting, Suggest edit, Emoji Reply
Invitee
Publish Note

Share your work with the world Congratulations! 🎉 Your note is out in the world Publish Note

Your note will be visible on your profile and discoverable by anyone.
Your note is now live.
This note is visible on your profile and discoverable online.
Everyone on the web can find and read all notes of this public team.
See published notes
Unpublish note
Please check the box to agree to the Community Guidelines.
View profile
Engagement control
Commenting
Permission
Disabled Forbidden Owners Signed-in users Everyone
Enable
Permission
  • Forbidden
  • Owners
  • Signed-in users
  • Everyone
Suggest edit
Permission
Disabled Forbidden Owners Signed-in users Everyone
Enable
Permission
  • Forbidden
  • Owners
  • Signed-in users
Emoji Reply
Enable
Import from Dropbox Google Drive Gist Clipboard
   owned this note    owned this note      
Published Linked with GitHub
Subscribed
  • Any changes
    Be notified of any changes
  • Mention me
    Be notified of mention me
  • Unsubscribe
Subscribe
# AdaBoost :::warning :notebook_with_decorative_cover: **摘要** [toc] ::: # 簡介 給定擁有 $N$ 個樣本的一組資料集: $$ \large D=\{(x_n,y_n)~~\forall~~n=1, ~\dots,~N\} $$ 其中 $$ \large y_n\in\{-1,~1\}~~\forall~~n=1,~\dots,~N $$ AdaBoost 的核心概念是依序建立 $T$ 個弱學習器 (weak learner) $h_t(x)$,每個模型會更關注在前個模型分類錯誤的那些樣本上。不僅如此,每個模型都將會被賦予一個模型權重 $\alpha_t$ ,該權重必須反映兩件事: - 模型權重越高,代表該模型效果較好。 - 模型權重越低,代表該模型效果較差。 如此一來,最終模型 $H(x)$ 就是蒐集這 $T$ 個弱學習器 (weak learner)的觀點,做出最終分類。也就是說,我們可以表示如下: :::success **重點** $$ \large H(x)=\text{sign}\Big(\sum_{t=1}^T \alpha_t h_t(x)\Big) $$ :::spoiler 其中 $$ \large h_t(x) \in \{-1,~1\}~~\forall~~t=1,~\dots,~T $$ ::: 值得注意的是,我們有提到這 $T$ 個模型是「依序」建立的,而且會更關注在前個模型「分類錯誤」的樣本上。換句話說,當前模型 $h_t(x)$ 的建立其實是依賴於前一個模型 $h_{t-1}(x)$,依此類推。 # 關注在「分類錯誤」的樣本上 至於該怎麼「更關注」在前個模型「分類錯誤」的樣本呢?概念其實很簡單: 首先,我們可以維護一組樣本權重,用來代表第 $t$ 個模型關注每個樣本的程度: :::success **重點** $$ \large W_{t}=\big[ w_{t1},~w_{t2},~\dots,~w_{tN} \big]^T $$ :::spoiler 其中 $$ \large 0<w_{tn}<1~~\forall~~n=1,~\dots,~N \\ ~ \\ \sum_{w~\in~W_t}w = 1 $$ ::: 接著,根據當前模型 $h_t(x)$ 的分類結果,更新當前樣本權重 $W_t$ 至 $W_{t+1}$。更新樣本權重的準則有兩個: - 「分類正確」的樣本必須給予更低的樣本權重。 - 「分類錯誤」的樣本必須給予更高的樣本權重。 如此一來,下個模型 $h_{t+1}(x)$ 就會更關注在那些前個模型分類錯誤的那些樣本上了。 # 如何計算模型權重? 目前我們已經知道每個模型 $h_t(x)$ 會有一個模型權重 $\alpha_t$,用來表示當前模型的重要程度。但是我們並不知道怎麼求得?實際上,我們可以透過以下步驟求得。 首先,我們定義一個用來衡量模型分類效果的指標(可理解為加權過後的分類錯誤率): :::success **重點** $$ \large \varepsilon_t = \sum_{n=1}^{N}w_{tn}\cdot I\big[h_t(x_n) \ne y_n\big] $$ :::spoiler 其中 $$ \large I\big[h_t(x_n) \ne y_n\big]= \begin{cases} 1 &\text{if}~~h_t(x_n) \ne y_n \\ 0 &\text{if}~~h_t(x_n) = y_n \end{cases} \\ ~\\ \large \varepsilon_t \le 0.5 $$ ::: 為什麼 $\varepsilon_t \le 0.5$ 呢?一個很直觀的理解是,若當前模型 $h_t$ 的 $\varepsilon_t \ge 0.5$,只要將當前模型的分類準則顛倒過來(過去分類為 $-1$ 的樣本就改為分類至 $1$,反之亦然),那麼新的 $\varepsilon_t$ 就會小於等於 $0.5$ 了。因此,以最小化加權分類錯誤率為目標的訓練方式,是不可能建立出 $\varepsilon_t \le 0.5$ 的模型的。 而模型權重 $\alpha_t$ 其實就只是將 $\varepsilon_t$ 經過某個函數轉換而已,定義如下: :::success **重點** $$ \large \alpha_t=\frac{1}{2}\log\Big(\frac{1-\varepsilon_t}{\varepsilon_t}\Big) $$ ::: 為什麼要這樣定義呢?請觀察以下這張圖,其中橫軸是 $\varepsilon_t$,縱軸則是 $\alpha_t$: ![](https://i.imgur.com/HVWv1Vg.png) 我們可以觀察到以下兩點: - 當模型分類能力就跟丟銅板沒兩樣($\varepsilon_t\approx0.5$)時,模型權重 $\alpha_t=0$,代表模型非常不值得參考,給予極低的模型權重。 - 當模型每猜必中($\varepsilon_t \approx 0$)時,模型權重 $\alpha_t\approx\infty$,代表模型非常值得參考,給予極高的模型權重。 如此一來,我們就得到符合我們預期的模型權重了。 **結論** - $\alpha_t$ 越高,代表該模型效果越好。 - $\alpha_t$ 越低,代表該模型效果越差。 # 如何更新樣本權重? 回憶一下,更新樣本權重的準則如下: - 「分類正確」的樣本必須給予更低的樣本權重。 - 「分類錯誤」的樣本必須給予更高的樣本權重。 實際上,給定第 $n$ 個樣本 $(x_n,~y_n)$,第 $t$ 個模型 $h_t$ 和第 $t$ 組樣本權重 $W_t$,更新公式可定義如下: :::success **重點** $$ \large w_{(t+1)n} := \frac{w_{tn} \cdot \exp \big(-\alpha_t y_n h_t(x_n)\big)}{\sum_{n=1}^N w_{(t+1)n}}\\ $$ ::: 分母就只是個 normalize 的計算,將 $W_{t+1}$ 的所有元素壓縮在 $0$ 到 $1$ ,並且總和為 $1$(全機率總和為 $1$)。重要的是分子的涵義。 針對 $\exp \big(-\alpha_t y_n h_t(x_n)\big)$ ,我們從兩個角度了解。 ### 分類正確 給定第 $n$ 個樣本 $(x_n,~y_n)$,第 $t$ 個模型 $h_t$,當模型分類正確時,$\exp \big(-\alpha_t y_n h_t(x_n)\big)$ 會等於 $\exp (-\alpha_t)$。將 $\varepsilon_t$ 畫在橫軸上,$\exp(-\alpha_t)$ 畫在縱軸上,圖形會長得像這樣。 ![](https://i.imgur.com/q0jHoRp.png) 我們可從圖中觀察到兩件事: - 當模型分類能力就跟丟銅板沒兩樣($\varepsilon_t\approx0.5$)時,$\exp(-\alpha_t) \approx 1$。因為模型分類能力非常不值得參考,就算當前樣本分類正確,更新時也不降低其樣本權重。 - 當模型每猜必中($\varepsilon_t \approx 0$)時,$\exp(-\alpha_t) \approx 0$。因為模型分類能力非常值得參考,只要當前樣本分類正確,更新時就大幅降低其樣本權重。 ### 分類錯誤 給定第 $n$ 個樣本 $(x_n,~y_n)$,第 $t$ 個模型 $h_t$,當模型分類錯誤時,$\exp \big(-\alpha_t y_n h_t(x_n)\big)$ 會等於 $\exp (\alpha_t)$。將 $\varepsilon_t$ 畫在橫軸上,$\exp(\alpha_t)$ 畫在縱軸上,圖形會長得像這樣。 ![](https://i.imgur.com/ykEMLuc.png) 我們可從圖中觀察到兩件事: - 當模型分類能力就跟丟銅板沒兩樣($\varepsilon_t\approx0.5$)時,$\exp(\alpha_t) \approx 1$。因為模型分類能力非常不值得參考,就算當前樣本分類錯誤,更新時也不降低其樣本權重。 - 當模型每猜必中($\varepsilon_t \approx 0$)時,$\exp(\alpha_t) \approx \infty$。因為模型分類能力非常值得參考,只要當前樣本分類錯誤,更新時就大幅提高其樣本權重。 ### 小結 以下是在不同情況下,怎麼更新樣本權重的準則: | 模型\分類 | 正確 | 錯誤 | | -------- | -------- | -------- | | **不值得參考** | 小幅度降低 | 小幅度提高 | | **值得參考** | 大幅度降低 | 大幅度提高| # 總結 以上就是我們 AdaBoost 的所有核心概念。接下來透過 pseudocode 從頭到尾再瞭解 AdaBoost 一遍吧。 ```python= N # 樣本個數 T # 分類器個數 # STEP 1: 初始化一組樣本權重 W = [1/N for _ in range(T)] # STEP 2: 建立 T 個分類器 for t in range(1, T+1): # STEP 2-1: 根據最佳化目標(最小化加權分類錯誤率 epsilon_t),訓練第 t 個分類器。 # STEP 2-2: 根據加權分類錯誤率 epsilon_t,計算模型權重 alpha_t。 # STEP 2-3: 根據模型權重 alpha_t 和分類結果,更新每個樣本權重 w_{tn}。 # STEP 3: 根據 T 個分類器和模型權重,給出最終分類結果。 ``` # 參考資料 - [AdaBoost, Clearly Explained, StatQuest with Josh Starmer](https://www.youtube.com/watch?v=LsK-xG1cLYA) - [Introduction to Adaptive Boosting, Hsuan-Tien Lin](https://www.csie.ntu.edu.tw/~htlin/course/ml08fall/doc/adaboost.pdf) - [機器學習: Ensemble learning之Bagging、Boosting和AdaBoost, Tommy Huang](https://medium.com/@chih.sheng.huang821/%E6%A9%9F%E5%99%A8%E5%AD%B8%E7%BF%92-ensemble-learning%E4%B9%8Bbagging-boosting%E5%92%8Cadaboost-af031229ebc3) ###### tags: `ML`

Import from clipboard

Paste your markdown or webpage here...

Advanced permission required

Your current role can only read. Ask the system administrator to acquire write and comment permission.

This team is disabled

Sorry, this team is disabled. You can't edit this note.

This note is locked

Sorry, only owner can edit this note.

Reach the limit

Sorry, you've reached the max length this note can be.
Please reduce the content or divide it to more notes, thank you!

Import from Gist

Import from Snippet

or

Export to Snippet

Are you sure?

Do you really want to delete this note?
All users will lose their connection.

Create a note from template

Create a note from template

Oops...
This template has been removed or transferred.
Upgrade
All
  • All
  • Team
No template.

Create a template

Upgrade

Delete template

Do you really want to delete this template?
Turn this template into a regular note and keep its content, versions, and comments.

This page need refresh

You have an incompatible client version.
Refresh to update.
New version available!
See releases notes here
Refresh to enjoy new features.
Your user state has changed.
Refresh to load new user state.

Sign in

Forgot password

or

By clicking below, you agree to our terms of service.

Sign in via Facebook Sign in via Twitter Sign in via GitHub Sign in via Dropbox Sign in with Wallet
Wallet ( )
Connect another wallet

New to HackMD? Sign up

Help

  • English
  • 中文
  • Français
  • Deutsch
  • 日本語
  • Español
  • Català
  • Ελληνικά
  • Português
  • italiano
  • Türkçe
  • Русский
  • Nederlands
  • hrvatski jezik
  • język polski
  • Українська
  • हिन्दी
  • svenska
  • Esperanto
  • dansk

Documents

Help & Tutorial

How to use Book mode

Slide Example

API Docs

Edit in VSCode

Install browser extension

Contacts

Feedback

Discord

Send us email

Resources

Releases

Pricing

Blog

Policy

Terms

Privacy

Cheatsheet

Syntax Example Reference
# Header Header 基本排版
- Unordered List
  • Unordered List
1. Ordered List
  1. Ordered List
- [ ] Todo List
  • Todo List
> Blockquote
Blockquote
**Bold font** Bold font
*Italics font* Italics font
~~Strikethrough~~ Strikethrough
19^th^ 19th
H~2~O H2O
++Inserted text++ Inserted text
==Marked text== Marked text
[link text](https:// "title") Link
![image alt](https:// "title") Image
`Code` Code 在筆記中貼入程式碼
```javascript
var i = 0;
```
var i = 0;
:smile: :smile: Emoji list
{%youtube youtube_id %} Externals
$L^aT_eX$ LaTeX
:::info
This is a alert area.
:::

This is a alert area.

Versions and GitHub Sync
Get Full History Access

  • Edit version name
  • Delete

revision author avatar     named on  

More Less

Note content is identical to the latest version.
Compare
    Choose a version
    No search result
    Version not found
Sign in to link this note to GitHub
Learn more
This note is not linked with GitHub
 

Feedback

Submission failed, please try again

Thanks for your support.

On a scale of 0-10, how likely is it that you would recommend HackMD to your friends, family or business associates?

Please give us some advice and help us improve HackMD.

 

Thanks for your feedback

Remove version name

Do you want to remove this version name and description?

Transfer ownership

Transfer to
    Warning: is a public team. If you transfer note to this team, everyone on the web can find and read this note.

      Link with GitHub

      Please authorize HackMD on GitHub
      • Please sign in to GitHub and install the HackMD app on your GitHub repo.
      • HackMD links with GitHub through a GitHub App. You can choose which repo to install our App.
      Learn more  Sign in to GitHub

      Push the note to GitHub Push to GitHub Pull a file from GitHub

        Authorize again
       

      Choose which file to push to

      Select repo
      Refresh Authorize more repos
      Select branch
      Select file
      Select branch
      Choose version(s) to push
      • Save a new version and push
      • Choose from existing versions
      Include title and tags
      Available push count

      Pull from GitHub

       
      File from GitHub
      File from HackMD

      GitHub Link Settings

      File linked

      Linked by
      File path
      Last synced branch
      Available push count

      Danger Zone

      Unlink
      You will no longer receive notification when GitHub file changes after unlink.

      Syncing

      Push failed

      Push successfully