周洺賢
  • NEW!
    NEW!  Connect Ideas Across Notes
    Save time and share insights. With Paragraph Citation, you can quote others’ work with source info built in. If someone cites your note, you’ll see a card showing where it’s used—bringing notes closer together.
    Got it
      • 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
      • Invite by email
        Invitee

        This note has no invitees

      • Publish Note

        Share your work with the world Congratulations! 🎉 Your note is out in the world Publish Note No publishing access yet

        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.

        Your account was recently created. Publishing will be available soon, allowing you to share notes on your public page and in search results.

        Your team account was recently created. Publishing will be available soon, allowing you to share notes on your public page and in search results.

        Explore these features while you wait
        Complete general settings
        Bookmark and like published notes
        Write a few more notes
        Complete general settings
        Write a few more notes
        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
      • Note Insights New
      • Engagement control
      • Make a copy
      • 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 Note Insights Versions and GitHub Sync Sharing URL Create Help
    Create Create new note Create a note from template
    Menu
    Options
    Engagement control Make a copy 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
  • Invite by email
    Invitee

    This note has no invitees

  • Publish Note

    Share your work with the world Congratulations! 🎉 Your note is out in the world Publish Note No publishing access yet

    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.

    Your account was recently created. Publishing will be available soon, allowing you to share notes on your public page and in search results.

    Your team account was recently created. Publishing will be available soon, allowing you to share notes on your public page and in search results.

    Explore these features while you wait
    Complete general settings
    Bookmark and like published notes
    Write a few more notes
    Complete general settings
    Write a few more notes
    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
    • Any changes
      Be notified of any changes
    • Mention me
      Be notified of mention me
    • Unsubscribe
    # 高等人工智慧 ***3/10*** ![](https://i.imgur.com/9kGlAjD.png) AI主要重要的部分分成三大類 1.學習 2.知識儲存記憶能力 3.推理及判斷之能力 目前AI的問題是在於轉換硬體的速度 現今的AI要做到行動像人都是以圖靈測試為基準 那要讓AI做到行動像人需要包含以下四大特性 1.自然語言表示 2.知識的表示(知識圖譜) 3.自動推論(貝氏推論技術) 4.機器學習 ![](https://i.imgur.com/8yZVGPF.png) ![](https://i.imgur.com/kbEEMs9.png) ***3/17 Agents*** ![](https://i.imgur.com/TM7mA9L.png) 這章主要討論理性行為 1.討論Agent的特質 2.評估環境的特質 3.Agent的類型 ![](https://i.imgur.com/QwMY4TN.png) 2.1 Agent總共有兩種行為一個是感測環境另一個則是行動 ![](https://i.imgur.com/7v0JItN.png) 在上圖中可得知有以一下三個Agent的概念 1.human Agent(人類):人類的感測是由眼睛、耳朵以及一些器官所組成的,那你的自動器則是以手、腿、聲音等由這些來去執行的。 2.robtic agent(機器人):機器人的感測器可以用攝影機或者是紅外線等來做感測,那它的行動則是透過各種馬達使機器人可以做出各種行動。 3.software agent(軟體):其實google的搜尋引擎就是屬於軟體的agent,通常這類的agent具有學習能力,它的感測來自於鍵盤、檔案或者事封包來進行感測。 感知會分為兩個概念: 1.感知一個點 2.感知一序列 ![](https://i.imgur.com/SG2E2I4.png) 我們要處理的問題是動態的話就是一個微分方程(電腦視覺就是動態) 我們要處理的問題是靜態的話就是一個函數關係 2.2什麼叫好的行為 ![](https://i.imgur.com/QOCOcrA.png) Rationality 1.性能量測的指標且可以定義在什麼情況下是成功 2.Agent內建環境的知識 3.我們設計的動作Agent可以完成 4.可以感知到最新的資料 理性只要是在當下是最好最佳就可以並不是要最安全 2-3 ![](https://i.imgur.com/ak56Ei6.png) 工作環境是一個很重要的指標 做AI一定要對你的題目做PEAS的了解 環境是全部可觀還是部份可觀 環境在設計上是確定性還是隨機性 如是隨機性就會較為複雜 但如果是不確定性就必須知道事件 靜態就是一個函數 動態的話是一個微分方程 考慮環境是靜態還是動態 靜態+時間的話就屬於半動態 你的環境是知道還是不知道 當一個專案的環境是一個不知道、是隨機的、需要多個Agent的等會是一個很困難情況 2-4 Agents=一個架構+程式 要讓查表(Lookup table)比較快就須做平行處理 ![](https://i.imgur.com/VUxPaU0.png) ![](https://i.imgur.com/xLnPkFl.png) 邏輯越強的話可解釋性Model就越大 Goal-based agents 他主要是在model-based agent 上的改進,所以它會有目標的資訊 Utility-based agents ![](https://i.imgur.com/UrHv76G.png) ![](ht5tps://i.imgur.com/8cSHbdy.png) Learning Agents ![](https://i.imgur.com/kZEFtEX.png) ***3/24*** Agent繼承關係圖 ![](https://i.imgur.com/jICMv6O.png) Muti-Agent是未來分散式AI的主要架構 做遊戲設計或是控制很常使用強化式學習 只要Agent架構裡面有reword(獎勵)的話就是強化式學習 加群平均是非線性的 ![](https://i.imgur.com/gc7kgcN.png) ***3/31*** ![](https://i.imgur.com/MIGCipa.png) 模糊系統在AI AI在處理不確定性問題主要是用貝式理論、模糊理論 弱AI的部分:雖然可以取代人部分且學習能力強但它可解釋度低,不可靠 強AI的部分: 圖神經網路與隨機森林可解釋度高也是現今比較熱議的 softmax層其實就是簡化版的模糊系統 可解釋性AI概念: 它可以藉由模糊系統或者是圖像AI等,讓使用者知道今天訓練出來的結果是以什麼為基準去得出的結果 ![](https://i.imgur.com/rs2xE0U.png) 如果我們要做模糊處理那輸入必須做模糊化 區間性的模糊集合是有上下性的 模糊集合只有中間式精準的其他是層度 投影可以算是模糊的降維 模糊運算他的交集(相加)取最大值聯集(相乘)取最小值 ![](https://i.imgur.com/68wRRjH.png) ![](https://i.imgur.com/j84WAOb.png) ![](https://i.imgur.com/OQykeqW.png) ![](https://i.imgur.com/pcvxDDA.png) ![](https://i.imgur.com/ug3BvJ3.png) ![](https://i.imgur.com/3sna7AO.png) ![](https://i.imgur.com/Gzm4qkG.png) Fuzzy整合可以從觀點1圖中裡的這四大領域做結合,且Fuzzy它就像我們人的左腦邏輯運算一樣使AI的思考可以更像人。 ![](https://i.imgur.com/fNnsoXf.png) Fuzzy logic缺點:是需要Knowledge base的提供,如果沒有專家去建立的話沒辦法做應用開發。 模糊推論系統(FIS): 1.每次輸入會是精準值須做模糊化變成模糊輸入。 2.做完模糊推論的規則會產出模糊集合輸出。 3.想要有精確的輸出必須做Defuzzifer。 4.在做建模糊化、推論工程、解模糊化需要有模糊知識庫。 知識建立: 1.將問題變成符號表示概念,也就是圖中IF then架構。 2.藉由此架構建立映射的關係 3.在建模時y會依解決問題來去表示當下示微分方程還是函數的關係 ![](https://i.imgur.com/x1SGXye.png) 模糊數:模糊數是針對模糊集合的某些動作做限制。 ![](https://i.imgur.com/b8r4Y11.png) 模糊集合方法: 1.模糊關係 2.模糊數 3.延伸定理 4.模糊最佳化 模糊測度方法: 1.模糊積分 模糊邏輯方法: 1.語言變數 2.模糊演算法 3.近似推論(在做模糊邏輯比較重要的地方) ![](https://i.imgur.com/v1QuIdT.png) 模糊集合如果是離散的話會像圖中底下那串供式表示。 模糊集合如果是連續型的則是會使用圖中底下的積分公式表示。 ![](https://i.imgur.com/YYDof2b.png) ![](https://i.imgur.com/9zfoI8O.png) ![](https://i.imgur.com/Fhh4rxq.png) ![](https://i.imgur.com/J8FfiAC.png) ![](https://i.imgur.com/vP7qgUk.png) 單條條rule:不用擔心有聚合的問題 多條rule:則需要擔心聚合、已經要取最大值或是最小值等問題 ![](https://i.imgur.com/GWsf4wJ.png) 高斯兩條rule的計算 ![](https://i.imgur.com/HIjIwTz.png) ![](https://i.imgur.com/ToFNFFG.png) 聚合概念 AND運算: 1.取最小值 2.rule是交集 OR運算: 1.取最大值 2.rule聯集 ![](https://i.imgur.com/76eaNrB.png) ![](https://i.imgur.com/wkNm5AC.png) ![](https://i.imgur.com/CExXXNK.png) 在TSK Fuzzy Model裡 z這個函數可以用線性、非線性或者是微分方程式來表示。 且此Model在做模糊整合較為簡單 ![](https://i.imgur.com/6XT2U7z.png) ![](https://i.imgur.com/KUa0tRl.png) 解模糊化可以用圖中5種做法 其中重心法COA運算較為複雜所以比較難做 ![](https://i.imgur.com/GSSCsAU.png) ![](https://i.imgur.com/pyw43FC.png) ***4/14*** ![](https://i.imgur.com/y8NYoqL.png) ![](https://i.imgur.com/yNxgJt6.png) ![](https://i.imgur.com/m2rUHnW.png) Type-2 Fuzzy 與前一代不同的是他比較一般化,且較為複雜可以處理較多的不確定性 Type-2要做解模化可能會要降成一階 聯集 maxmum ![](https://i.imgur.com/XgOCD8s.png) 4/28 ***NLP(自然語言處理&神經網路語言處理)*** ![](https://i.imgur.com/2j8lVar.png) ![](https://i.imgur.com/UqjsWem.png) ![](https://i.imgur.com/Hgzfji9.png) ![](https://i.imgur.com/0PQmn4Y.png) NLP=NLU(語言理解)+NLG(語言產生) ![](https://i.imgur.com/yLQwVS4.png) ![](https://i.imgur.com/LR2NVDd.png) ![](https://i.imgur.com/g4WNX8H.png) ![](https://i.imgur.com/V3pdUPK.png) ![](https://i.imgur.com/mlOjBkz.png) ![](https://i.imgur.com/l7MyBed.png) 向量:200~400 ***強化學習*** 強化式學習就是屬於監督式還是非監督式學習這兩種則中去做(例:離散學習也算是強化式學習) ![](https://i.imgur.com/C37SXIr.png) 強化式學習可以分成兩大類分別為: 1.MDP已知 模型會使用動態規劃分以下兩種: (1).策略迭帶 (2).值迭帶 可以藉由以上兩種迭帶方法讓model知道目前訓練出來的結果好不好 2.MDP未知 馬可夫決策過程是隨機(不確定性)的輸入模式所以須使用MDP 模型藉由以下兩種建表來得出最好的結果: (1.)蒙地卡羅方法 (2.)時序差分 但這種方法無法知道決策過程 ![](https://i.imgur.com/8CsDW6u.png) Model based 方法取樣較快原因是建立在Model是在已知決策過程中條件下取樣效率相對應較快 Off-policy概念:把學習過程用表格建起來,或者用模擬器嘗試把未知的東西用表格建起來,也因為這原因在用這種方法做學習model取樣效率也會比On-policy來的好 ![](https://i.imgur.com/iwftYtp.png) ![](https://i.imgur.com/livQcx0.png) ![](https://i.imgur.com/ITO5gt2.png) Agent會透過以下三種跟環境去做連結與變化並給予相對應的獎勵或評斷 1.Observation:觀察當下的狀態 2.Action:對當下的狀態去做策略 3.Reward:可以透過這函數去做評論 ![](https://i.imgur.com/FfTpolB.png) ![](https://i.imgur.com/o66Ovvq.png) ![](https://i.imgur.com/5X5dWek.png) ![](https://i.imgur.com/KMLjc19.png) ![](https://i.imgur.com/abNpDa6.png) 在強化式學習裡獎勵函數可分成Rewards、Returns Rewards、Returns兩者差異: Rewards:它每次獎勵的內容會隨則次數增加而將獎勵內容逐漸變少(比較看重第一步好不好) Returns:它每獎勵的內容都一致不會有獎勵變少問題 ![](https://i.imgur.com/xsQgeP1.png) State-value:是將Action做期望值 Optimal action-value:是在取最大值 ![](https://i.imgur.com/pAZs8ID.png) ![](https://i.imgur.com/LLctzbj.png) ![](https://i.imgur.com/zeX1qMM.png) ![](https://i.imgur.com/DD8UTA9.png) ![](https://i.imgur.com/lbP3TPy.png) ![](https://i.imgur.com/abwsWqW.png) ![](https://i.imgur.com/1LP1Zzc.png) ![](https://i.imgur.com/JBfWnfM.png) ![](https://i.imgur.com/Ba24Gs4.png) ![](https://i.imgur.com/dPMgSuy.png) ![](https://i.imgur.com/TE22vvm.png) ![](https://i.imgur.com/tS0eO7r.png) ![](https://i.imgur.com/SdqMaqp.png) ![](https://i.imgur.com/dRyMCZd.png) ![](https://i.imgur.com/3UxTJkh.png) ![](https://i.imgur.com/bkeyZyb.png) ![](https://i.imgur.com/9eUcgZ4.png) ![](https://i.imgur.com/xKaEEdt.png) ![](https://i.imgur.com/sMZUb6O.png) ![](https://i.imgur.com/mExcGRf.png) ![](https://i.imgur.com/IoSUxM1.png) ![](https://i.imgur.com/59M2W0f.png) ![](https://i.imgur.com/vxVi5Zz.png) ![](https://i.imgur.com/uRr0pqx.png) ![](https://i.imgur.com/0kULBF2.png) ![](https://i.imgur.com/vm4b7Pk.png) ![](https://i.imgur.com/VnQw5Xj.png) ![](https://i.imgur.com/hXbm0Gb.png) ![](https://i.imgur.com/uswryNO.png) ![](https://i.imgur.com/5q7NQVP.png) ![](https://i.imgur.com/Jzyp7f2.png) ![](https://i.imgur.com/L9ON11x.png) --- ## 開放教育平台-影片(簡短重點) 物件偵測主要目的在於可以定位物體和辨識物體,那物件偵測又分為2大偵測方式 1. Two stages detection(二段式偵測) 2. single stage detection(一段式偵測) 二段式偵測主要偵測速度慢但是辨識物體準確度高 一段式偵測主要偵測速度快但是辨識物體準確度相對二段式偵測低 ![](https://i.imgur.com/xl7CaDI.png) ### 演算法 一段式偵測的物件偵測演算法的代表為 Yolo (You Only Look Once)和SSD... 二段式偵測為 1.R-CNN、2.Fast R-CNN、3.Fasted R-CNN和4. R-FCN ... ### 評估模型好壞的指標 ![](https://i.imgur.com/12l4fai.png) 1.TP(True Positive)、2. FP(False Positive)、3. TN(True Negative)和4. FN(False Negative)。此外也利用這4種指標延伸出了2種新指標為Precision(精確度)和recall(召回率) ![](https://i.imgur.com/QwCUsRn.png) ### Training(訓練)和Inference(推論)的不同 ![](https://i.imgur.com/DWfOd0o.png) ### 深度學習的模型壓縮和加速 最常見的方法為Purning(裁剪)。那Purning有許多種方式。 1. weighting Purning(權重裁剪) 2. Layer Purning(層數裁剪) 3. Channel Purning(通道裁剪)等許多方式進行壓縮 ![](https://i.imgur.com/3dSGMFy.png) ![](https://i.imgur.com/fmdgUxK.png)

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

    New to HackMD? Sign up

    By signing in, you agree to our terms of service.

    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