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.
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
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!
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?
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.
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.
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
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?
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
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?
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?
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)
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
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.
Thanks for your comments and feedback!
This is the best implementation of SOMs I have seen.
Thanks for comments!
Siddharth Yadav agreed. This guy is awesome!
@@Jerry-uc1pn Thanks!
Agreed too! Thank you for uploading this.
Thanks for comments!
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
Glad you like them!
@@bkrai do you have a video on LSTM please in R ?
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.
@@bkrai cant wait. Make it a xmas gift
Good idea :)
Thank you so much. This video helped to figure out how to use the Kohonen package on meteorological data.
You are very welcome!
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!
For nominal variables, the typical encoding in a neural network is one-hot encoding.
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?
Great tutorial Sir. Thanks a lot.
Thanks for comments!
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.
Very good. Do you think this technique is useful for reducing the dimensionality of mixed data (quantitative variables and factors)?
Yes, absolutely
Great tutorial! Thanks!
Thanks for comments!
Thank you for the great explanation! However, I was wondering how exactly are the labels assigned to the units?
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.
Hi Sir, could please tell a bit on why you normalized test_x based on train dataset distribution
Great Video - is it possible to use it for multi-label text classification?
Yes, it should work.
Amazing video. I get go ahead with this video.
Thanks for feedback!
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.
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.
Dr. Bharatendra Rai thank you. Well appreciated
welcome!
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
Thanks for the update!
Thank you! Any function or procedure useful to decide how many nodes to use in the unsupervised case?
It's mostly depends on type of data and also involves trial and error.
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?
You can refer to this:
th-cam.com/video/0xsM0MbRPGE/w-d-xo.html
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
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?
I used it as an example, but you can try other grid sizes as well.
Great video. Could please teach us about NLP in R? Regards
You can use this link:
th-cam.com/play/PL34t5iLfZddt0tt5GdDy3ny6X5RQvwrp6.html
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?
thank you for sharing. btw how to know accuracy using som for classification ? thanks
I showed how to find accuracy using som for classification in the video.
@@bkrai thank you. btw how do we know learning rate used in that classification sir ? is 0.05 ? thanks
When he does the supervised SOM for classification & prediction, is that the same thing as LVQ or different? Thanks
Yes.
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
Very nice
Thanks for comments!
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)
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
Seeing this today. Probably you have already figured out.
Sir, how can we do the supervised clustering of continous response variable?
Dear Bharatendra Rai
In multiple imputation, how to decide on which the best proposed from 3 or 5 imputation?
probably you posted your question on wrong video.
THANKS
Can we do clustering on the non-supervise method? If yes, how?
It shows steps for both supervised and unsupervised situations.
@@bkrai right lol, but I mean how do we know how many clusters there are, like what is the value for k?
That depends on data and nature of the business problem, and can involve trial and error approach.
@@bkrai okay, thank you!
Welcome!
Can we used as dimensionality reduction technique.
Yes.
speak about adaline, i am very dificulty in programing this.
Thanks for the suggestion, I've added this to my list.
+Bharatendra Rai thanks 😊
Good video. My friend how i can see the names of th objects in the map?
Something similar to this -> $unit.classif
For example... In the circle 1 what is the name of the objects that are there?
Sir, how we know weighted in SOM ? Thanks
Which part or stage of the video you are referring to?
I can't get kohonen library installed. Any suggestions?
I would suggest, try again. I was able to install it from RStudio without any problem.
it could be your r version is out of date , checked the news update version in both r and your rstudio
Nice, Can you upload data please
Here is the link:
drive.google.com/open?id=0B5W8CO0Gb2GGVjRILTdWZkpJU1E
unsupervised 1.13
Thanks