Self Organizing Maps in R | Kohonen Networks for Unsupervised and Supervised Maps

แชร์
ฝัง
  • เผยแพร่เมื่อ 4 พ.ย. 2024

ความคิดเห็น • 89

  • @ramram2utube
    @ramram2utube 4 ปีที่แล้ว +1

    Oh, I was searching for exactly this type of explanation only for implementing SOM using R and I could get it. Many thanks to Dr. Bharatendra Rai. Now I should be ready for tomorrow's training lecture to be taken on SOM more confidently!!. Such lectures are really help us get a clear understanding of what is going behind constructing SOM maps in a lucid manner. I appreciate Dr. Rai for his lecture which has both depth and clarity. As a statistician, I appreciate the time and efforts devoted by Dr. Rai for explaining the whole topic via R. I am grateful to him for providing the same on youtube for the benefit of all.

    • @bkrai
      @bkrai  4 ปีที่แล้ว

      Thanks for your comments and feedback!

  • @siddharthyadav9734
    @siddharthyadav9734 6 ปีที่แล้ว +9

    This is the best implementation of SOMs I have seen.

    • @bkrai
      @bkrai  6 ปีที่แล้ว +1

      Thanks for comments!

    • @Jerry-uc1pn
      @Jerry-uc1pn 4 ปีที่แล้ว +1

      Siddharth Yadav agreed. This guy is awesome!

    • @bkrai
      @bkrai  3 ปีที่แล้ว +1

      @@Jerry-uc1pn Thanks!

    • @程曦-r6o
      @程曦-r6o 3 ปีที่แล้ว +1

      Agreed too! Thank you for uploading this.

    • @bkrai
      @bkrai  3 ปีที่แล้ว

      Thanks for comments!

  • @alphar85
    @alphar85 3 ปีที่แล้ว

    Rai, you probably see my comments in all of your videos :) I just wanna thank you for all these videos man. I wish I came across or put the effort to find them few years back. you are genuinely the best I have seen on TH-cam. such a blessing

    • @bkrai
      @bkrai  3 ปีที่แล้ว

      Glad you like them!

    • @alphar85
      @alphar85 3 ปีที่แล้ว

      @@bkrai do you have a video on LSTM please in R ?

    • @bkrai
      @bkrai  3 ปีที่แล้ว

      I've a chapter in my 'advanced deep learning with R' book, but not video. Hopefully next month or so I'll do a video too.

    • @alphar85
      @alphar85 3 ปีที่แล้ว +1

      @@bkrai cant wait. Make it a xmas gift

    • @bkrai
      @bkrai  3 ปีที่แล้ว

      Good idea :)

  • @CoMtLion
    @CoMtLion 3 ปีที่แล้ว +1

    Thank you so much. This video helped to figure out how to use the Kohonen package on meteorological data.

    • @bkrai
      @bkrai  3 ปีที่แล้ว +1

      You are very welcome!

  • @srikalachitti890
    @srikalachitti890 2 ปีที่แล้ว +2

    Hello Dr.Bharatendra, you gave a very clear explanation and interpretation of SOMs. Can we use SOMs for nominal data? If yes, should we do it with the scaling of the data? Thanks in advance!

    • @bkrai
      @bkrai  2 ปีที่แล้ว

      For nominal variables, the typical encoding in a neural network is one-hot encoding.

  • @arowoloolatunbosun1368
    @arowoloolatunbosun1368 2 ปีที่แล้ว

    Thanks, Dr. Bharatendra Rai for this video. In generating the SOM for distant neighbours, is the colour assigned similar to that seen in the count map? In addition, is the arrangement of the count nodes also a way to show the level of relatedness e.g node 1 and node 2 somewhat related because they are close to each other?

  • @flamboyantperson5936
    @flamboyantperson5936 6 ปีที่แล้ว +2

    Great tutorial Sir. Thanks a lot.

    • @bkrai
      @bkrai  6 ปีที่แล้ว

      Thanks for comments!

  • @rupshanandi2730
    @rupshanandi2730 5 หลายเดือนก่อน

    Sir, thank you so much for this amazing lecture. I would be really grateful if you could help me a liitle more regarding this. Can you please suggest me some research articles where similar predictions have been performed using SOM.

  • @sebastianvarela2190
    @sebastianvarela2190 4 ปีที่แล้ว +1

    Very good. Do you think this technique is useful for reducing the dimensionality of mixed data (quantitative variables and factors)?

    • @bkrai
      @bkrai  4 ปีที่แล้ว

      Yes, absolutely

  • @1novisad
    @1novisad 6 ปีที่แล้ว +1

    Great tutorial! Thanks!

    • @bkrai
      @bkrai  6 ปีที่แล้ว

      Thanks for comments!

  • @drummerAJ123
    @drummerAJ123 4 ปีที่แล้ว +1

    Thank you for the great explanation! However, I was wondering how exactly are the labels assigned to the units?

    • @bkrai
      @bkrai  4 ปีที่แล้ว

      I would suggest going through some book that covers algorithm in more detail. In this video the attempt was to illustrate steps for doing it in R.

  • @MahbuburRahman-ro2xy
    @MahbuburRahman-ro2xy ปีที่แล้ว

    Hi Sir, could please tell a bit on why you normalized test_x based on train dataset distribution

  • @suryaprakashb1629
    @suryaprakashb1629 5 ปีที่แล้ว +2

    Great Video - is it possible to use it for multi-label text classification?

    • @bkrai
      @bkrai  4 ปีที่แล้ว

      Yes, it should work.

  • @gustavoarco
    @gustavoarco 5 ปีที่แล้ว +1

    Amazing video. I get go ahead with this video.

    • @bkrai
      @bkrai  5 ปีที่แล้ว

      Thanks for feedback!

  • @ndiyekebonye208
    @ndiyekebonye208 4 ปีที่แล้ว +1

    Thank you Dr Rai for the tutorial but i am quite interested in how you ascertained the xdim and ydim at the beginning of the video. How did you know these were the best dimensions of the grid for the SEOM. Thank you.

    • @bkrai
      @bkrai  4 ปีที่แล้ว +2

      By mostly trial and error. It's difficult to say they are the best and only way. Sometimes given the data, it could also be your personal choice.

    • @ndiyekebonye208
      @ndiyekebonye208 4 ปีที่แล้ว +1

      Dr. Bharatendra Rai thank you. Well appreciated

    • @bkrai
      @bkrai  4 ปีที่แล้ว +1

      welcome!

    • @ndiyekebonye208
      @ndiyekebonye208 4 ปีที่แล้ว +1

      Dr. Bharatendra Rai i found a way to solve it. I came up with various hypothetical maps after using 5 times square root of sample size. Then compared average quantization errors of each map. The one with the least error was settled for. Thanks

    • @bkrai
      @bkrai  4 ปีที่แล้ว

      Thanks for the update!

  • @juansb1509
    @juansb1509 6 ปีที่แล้ว +2

    Thank you! Any function or procedure useful to decide how many nodes to use in the unsupervised case?

    • @bkrai
      @bkrai  6 ปีที่แล้ว

      It's mostly depends on type of data and also involves trial and error.

  • @rupsayar
    @rupsayar ปีที่แล้ว +1

    Hi, if I want to create and store a table with nodes (V1, V2, ...) as columns and applicant no. (assigned to their respective node) in rows, what is the code for the same?

    • @bkrai
      @bkrai  ปีที่แล้ว

      You can refer to this:
      th-cam.com/video/0xsM0MbRPGE/w-d-xo.html

  • @katherineherlambang6163
    @katherineherlambang6163 5 ปีที่แล้ว

    Hallo, thank you for the detailed explanation. But, how we know about the weight for each kriteria? for example in the video, how we know the weight of gre, gpa, and rank? thankyou

  • @Kosxd
    @Kosxd 3 ปีที่แล้ว +1

    Why u decided to use 4x4 grid with 16 neurons? As far as I read about it - it says grid size = 5xsqrt(number of samples). So in your case 5x20 = 100
    Where I'm wrong?

    • @bkrai
      @bkrai  3 ปีที่แล้ว

      I used it as an example, but you can try other grid sizes as well.

  • @Didanihaaaa
    @Didanihaaaa 6 ปีที่แล้ว +2

    Great video. Could please teach us about NLP in R? Regards

    • @bkrai
      @bkrai  6 ปีที่แล้ว +1

      You can use this link:
      th-cam.com/play/PL34t5iLfZddt0tt5GdDy3ny6X5RQvwrp6.html

  • @nuijyl6738
    @nuijyl6738 3 หลายเดือนก่อน

    Hi i have encountered a problem. I tired this algorithm multiple time but the prediction outcome i got is not always the same provided the same seed and parameters. Mya i know is this normal? Why this happen and how to solve it?

  • @putriyula
    @putriyula 5 ปีที่แล้ว +2

    thank you for sharing. btw how to know accuracy using som for classification ? thanks

    • @bkrai
      @bkrai  5 ปีที่แล้ว +2

      I showed how to find accuracy using som for classification in the video.

    • @putriyula
      @putriyula 5 ปีที่แล้ว

      @@bkrai thank you. btw how do we know learning rate used in that classification sir ? is 0.05 ? thanks

  • @indyd9322
    @indyd9322 3 ปีที่แล้ว +1

    When he does the supervised SOM for classification & prediction, is that the same thing as LVQ or different? Thanks

    • @bkrai
      @bkrai  2 ปีที่แล้ว

      Yes.

  • @abitharam8200
    @abitharam8200 4 ปีที่แล้ว

    Sir could you plz tell me how to validate the SOM cluster in R? Clvalid function works for other clustering method and not for som

  • @vjysri2756
    @vjysri2756 4 ปีที่แล้ว +1

    Very nice

    • @bkrai
      @bkrai  4 ปีที่แล้ว

      Thanks for comments!

  • @piotrratajczak4582
    @piotrratajczak4582 4 ปีที่แล้ว

    Thank you for your great and inspiring work! :) Could you please help me and give some tips on how to design trajectory/line (not map) using a one-dimensional Kohonen map? I need it to my article on bankruptcy but as I'm not statistician I can't handle this. Best, Piotr (Poland)

  • @fin101-z3o
    @fin101-z3o 5 ปีที่แล้ว +1

    thank you for the video. am new to R, when i entered the command plot(map, type = 'changes') the diagram became too big i couldnt see the numbers on the x and y axis. How do i fix that? thanks

    • @bkrai
      @bkrai  2 ปีที่แล้ว

      Seeing this today. Probably you have already figured out.

  • @praveenparmar7728
    @praveenparmar7728 5 ปีที่แล้ว

    Sir, how can we do the supervised clustering of continous response variable?

  • @larbihouichi8942
    @larbihouichi8942 5 ปีที่แล้ว

    Dear Bharatendra Rai
    In multiple imputation, how to decide on which the best proposed from 3 or 5 imputation?

    • @bkrai
      @bkrai  5 ปีที่แล้ว

      probably you posted your question on wrong video.

  • @sk93359
    @sk93359 6 ปีที่แล้ว +1

    THANKS

  • @vivianbalagon8281
    @vivianbalagon8281 4 ปีที่แล้ว +1

    Can we do clustering on the non-supervise method? If yes, how?

    • @bkrai
      @bkrai  4 ปีที่แล้ว

      It shows steps for both supervised and unsupervised situations.

    • @vivianbalagon8281
      @vivianbalagon8281 4 ปีที่แล้ว +1

      @@bkrai right lol, but I mean how do we know how many clusters there are, like what is the value for k?

    • @bkrai
      @bkrai  4 ปีที่แล้ว +1

      That depends on data and nature of the business problem, and can involve trial and error approach.

    • @vivianbalagon8281
      @vivianbalagon8281 4 ปีที่แล้ว +1

      @@bkrai okay, thank you!

    • @bkrai
      @bkrai  4 ปีที่แล้ว

      Welcome!

  • @osamaaljarrah9851
    @osamaaljarrah9851 4 ปีที่แล้ว +1

    Can we used as dimensionality reduction technique.

    • @bkrai
      @bkrai  4 ปีที่แล้ว

      Yes.

  • @VamosCoringar
    @VamosCoringar 6 ปีที่แล้ว +1

    speak about adaline, i am very dificulty in programing this.

    • @bkrai
      @bkrai  6 ปีที่แล้ว

      Thanks for the suggestion, I've added this to my list.

    • @VamosCoringar
      @VamosCoringar 6 ปีที่แล้ว +1

      +Bharatendra Rai thanks 😊

  • @xlejxndromecixs5695
    @xlejxndromecixs5695 5 ปีที่แล้ว

    Good video. My friend how i can see the names of th objects in the map?

    • @xlejxndromecixs5695
      @xlejxndromecixs5695 5 ปีที่แล้ว

      Something similar to this -> $unit.classif
      For example... In the circle 1 what is the name of the objects that are there?

  • @putriyula
    @putriyula 5 ปีที่แล้ว +1

    Sir, how we know weighted in SOM ? Thanks

    • @bkrai
      @bkrai  5 ปีที่แล้ว

      Which part or stage of the video you are referring to?

  • @sudhanshuchadha
    @sudhanshuchadha 6 ปีที่แล้ว

    I can't get kohonen library installed. Any suggestions?

    • @bkrai
      @bkrai  6 ปีที่แล้ว +1

      I would suggest, try again. I was able to install it from RStudio without any problem.

    • @colinchan8536
      @colinchan8536 4 ปีที่แล้ว

      it could be your r version is out of date , checked the news update version in both r and your rstudio

  • @sudhanshuchadha
    @sudhanshuchadha 6 ปีที่แล้ว +1

    Nice, Can you upload data please

    • @bkrai
      @bkrai  6 ปีที่แล้ว

      Here is the link:
      drive.google.com/open?id=0B5W8CO0Gb2GGVjRILTdWZkpJU1E

  • @bellarahmatullah2012
    @bellarahmatullah2012 4 ปีที่แล้ว +1

    unsupervised 1.13

    • @bkrai
      @bkrai  4 ปีที่แล้ว

      Thanks