Close Menu
    Trending
    • AMD CEO Claims New AI Chips ‘Outperform’ Nvidia’s
    • How AI Agents “Talk” to Each Other
    • Creating Smart Forms with Auto-Complete and Validation using AI | by Seungchul Jeff Ha | Jun, 2025
    • Why Knowing Your Customer Drives Smarter Growth (and Higher Profits)
    • Stop Building AI Platforms | Towards Data Science
    • What If Your Portfolio Could Speak for You? | by Lusha Wang | Jun, 2025
    • High Paying, Six Figure Jobs For Recent Graduates: Report
    • What If I had AI in 2018: Rent the Runway Fulfillment Center Optimization
    Finance StarGate
    • Home
    • Artificial Intelligence
    • AI Technology
    • Data Science
    • Machine Learning
    • Finance
    • Passive Income
    Finance StarGate
    Home»Machine Learning»Xaier Initialization 神經網路參數初始化 – Jacky Chou
    Machine Learning

    Xaier Initialization 神經網路參數初始化 – Jacky Chou

    FinanceStarGateBy FinanceStarGateMarch 12, 2025No Comments1 Min Read
    Share Facebook Twitter Pinterest LinkedIn Tumblr Reddit Telegram Email
    Share
    Facebook Twitter LinkedIn Pinterest Email


    Jacky Chou

    今天簡單介紹Xaier Initialization這個方法
    這個方法的目的是在初始化神經網路參數的時候穩定梯度
    避免梯度爆炸或是梯度消失
    可以看到下方算式代表的是一層神經網路參數W乘以輸入x得到輸出y
    W是矩陣 x是向量
    n_in n_out代表輸入的維度跟輸出的維度

    我們的目標就是讓輸出的變異數等於輸入的變異數(見第二個算式)

    藉此來穩定我們signal 傳遞

    我們假設W跟x是獨立變數 所以變異數就可以拆開變成相乘
    接下來W代表初始化根據正態分布 均值0 變異數sigma平方
    x則是輸入 輸入通常都會正則化 代表均值0 變異數1
    接著帶入rearange得出sigma平方=1/_in

    接著可以看到下面第一行我們要將損失函數做一個反向傳播的動作
    那這邊我們關注的就會是損失函數梯度變異數是否跟輸入的變異數一樣
    我們用chain rule來反向到xi 輸入層
    第二行就是對xi 偏微分其他變成0 剩下Wji
    接下來對損失函數梯度指定它為小delta
    套上變異數 跟昨天一樣 輸入跟神經網路參數獨立
    所以可以變成相乘
    接下來解開sigma
    得到結果

    你會發現說正向傳播跟反向傳播怎麼一個分母是輸入維度
    一個是輸出維度 那無法同時滿足兩個要求
    所以就有學者提出第二張最終的公式
    參數wji 就會從正態分布均值0標準差為根號變異數當中提取數字



    Source link

    Share. Facebook Twitter Pinterest LinkedIn Tumblr Email
    Previous ArticleFeature Comparison: Leading RPA Workflow Tools
    Next Article Heatmaps for Time Series  | Towards Data Science
    FinanceStarGate

    Related Posts

    Machine Learning

    Creating Smart Forms with Auto-Complete and Validation using AI | by Seungchul Jeff Ha | Jun, 2025

    June 14, 2025
    Machine Learning

    What If Your Portfolio Could Speak for You? | by Lusha Wang | Jun, 2025

    June 14, 2025
    Machine Learning

    YouBot: Understanding YouTube Comments and Chatting Intelligently — An Engineer’s Perspective | by Sercan Teyhani | Jun, 2025

    June 13, 2025
    Add A Comment

    Comments are closed.

    Top Posts

    09360627233

    March 26, 2025

    Alberta-based Corinne, 69, wonders if her retirement savings will last

    April 9, 2025

    Exporting MLflow Experiments from Restricted HPC Systems

    April 24, 2025

    Grab Microsoft Office Professional Plus 2019 for Windows While It’s Just $30

    April 23, 2025

    The Precarious Life of Dual Unemployed Parents (DUPs)

    April 11, 2025
    Categories
    • AI Technology
    • Artificial Intelligence
    • Data Science
    • Finance
    • Machine Learning
    • Passive Income
    Most Popular

    GitHub Models 免費試玩 GPT、Llama、DeepSeek | by Tsai Yi Lin | Feb, 2025

    February 2, 2025

    Linear Regression | Basic Intuition | by techwithsujith | Apr, 2025

    April 23, 2025

    New AI tool generates realistic satellite images of future flooding | MIT News

    February 16, 2025
    Our Picks

    Making a fast RL env in C with pufferlib | by BoxingBytes | Mar, 2025

    March 27, 2025

    AI Engineering (3/3): Dataset Engineering, Inference Optimization, and Architecture and User Feedback | by Marina Wyss – Gratitude Driven | Data Science Collective | Mar, 2025

    March 22, 2025

    AI-enabled control system helps autonomous drones stay on target in uncertain environments | MIT News

    June 9, 2025
    Categories
    • AI Technology
    • Artificial Intelligence
    • Data Science
    • Finance
    • Machine Learning
    • Passive Income
    • Privacy Policy
    • Disclaimer
    • Terms and Conditions
    • About us
    • Contact us
    Copyright © 2025 Financestargate.com All Rights Reserved.

    Type above and press Enter to search. Press Esc to cancel.