+ Post New Thread
Results 1 to 4 of 4
  1. #1
    Junior Member level 2
    Points: 570, Level: 5

    Join Date
    Nov 2014
    Posts
    20
    Helped
    0 / 0
    Points
    570
    Level
    5

    What is meant by Manhattan update rule in neural networks ?

    Could anyone explain me about

    1. What is meant by Manhattan update rule ?

    2. How to use Manhattan update rule in delta rule ?

    Thank you.

    •   Alt3rd July 2015, 06:41

      advertising

        
       

  2. #2
    Advanced Member level 5
    Points: 34,060, Level: 45
    Achievements:
    7 years registered Created Blog entry Created Album pictures
    SunnySkyguy's Avatar
    Join Date
    Sep 2007
    Location
    Richmond Hill, ON, Canada
    Posts
    6,600
    Helped
    1652 / 1652
    Points
    34,060
    Level
    45
    Blog Entries
    7

    Re: What is meant by Manhattan update rule in neural networks ?

    Been there.. But never tried it ;)

    http://www.helwan.edu.eg/university/...dnn_crf_MH.pdf

    it seems no delta on estimate just polarity.
    To converge the estimate STEP SIZE is incremented each time.

    seems like a simple but slow method but resilient

    like dynamic RAM Virtual Memory size reserve estimates.... Adapt with margin , grow by increasing larger steps at some rate until converge on desired value

    , stable, resilient , wide range

    not sure if step size reset to base increment on polarity toggle.. Probably
    A good design question lists your overall requirements™ The best question deserves a better answer. ™
    ... Tony Stewart EE since 1975



    •   Alt3rd July 2015, 07:05

      advertising

        
       

  3. #3
    Junior Member level 2
    Points: 570, Level: 5

    Join Date
    Nov 2014
    Posts
    20
    Helped
    0 / 0
    Points
    570
    Level
    5

    Re: What is meant by Manhattan update rule in neural networks ?

    Even I too read that Pdf. But I didnt understand. If possible, can you explain me in short about Manhattan update rule.



    •   Alt3rd July 2015, 07:09

      advertising

        
       

  4. #4
    Advanced Member level 5
    Points: 34,060, Level: 45
    Achievements:
    7 years registered Created Blog entry Created Album pictures
    SunnySkyguy's Avatar
    Join Date
    Sep 2007
    Location
    Richmond Hill, ON, Canada
    Posts
    6,600
    Helped
    1652 / 1652
    Points
    34,060
    Level
    45
    Blog Entries
    7

    Re: What is meant by Manhattan update rule in neural networks ?

    Ok , we both learned something
    A good design question lists your overall requirements™ The best question deserves a better answer. ™
    ... Tony Stewart EE since 1975



--[[ ]]--