Thank you for this walkthrough! It is very well done. I was looking every where to find an example of how to update the cluster matrix and this really helps. It is so well done and is extremely clear. Thank you for this series on clustering.
I would like to thank you for this video. Your explanation is magnificent and so clearly. You helped me a lot to comprehend these complex subjects. Greetings from Mexico.
Hello, Thanks a lot for the simple and clear explanation for the single linkage (previous video) and for the complete linkage as well. I have two questions. 1) Looking at the dendrograms obtained from the single linkage and the complete linkage, one can see that they are different. So , how can we interpret that? If I cut the tree at the same value (i mean for the single and complete linkage trees), I would obtain different clusters results. 2) What is the idea behind searching for the maximum distance in case of complete linkage?
Thank you Maam. That was a clean video and helped me a lot understanding the Complete-Link. I have a few questions question.. 1. How does the merge criterion influence the merge decision? 2. Why is this complete link clustering called non-local while the single link criterion called local?
Very good explanations! Can you please Show an example how to Use the correlation matrix as a distance matrix in kmeans. You have applied the euclidian distance in k means to cluster. How does the calculation of the Clusters work, with taking not the original dataset, but having the correlation matrix. How to Use the corr Matrix to Bild k means clusters? Thank you!
Thanks a ton for a fantastic explanation madam! When we pick to first start the merging process, shouldn't we pick P6 and P5 to merge first since it has Max value 0.39?
The first cluster is determined through the most similar units. After that we define the distance from that cluster to the other data points through either single linkage (looking at minimal distance) or complete link (looking at maximum distance)
COMPLETE LINK - it means, while calculating distance matrix, we take the maximum value, right? SINGLE LINK - while calculating, distance matrix, we take the minimum value?
Can you please explain what to do when the matrix has two same low value (eg: If P2 and P1 has 0.12 and P3 and P4 has 012). In that case which points need to be considered?
Try to upload classification sums naive Bayes ,bayessian and id3 ur video is very help full mumbai university students. ..try to solve it in same method followed by them .....Thanks
dear mam, i would like to know that sometimes cases appear where after computing the similarity matrix we find two lowest distances . Now we can choose anyone of the distances to merge at that step. Now this decision may affect the cluster output at the final stage . Well here am talking about the case when a distance threshold is applied . say for eg-{1,2,3,4,9,8,7}.here if we take a threshold of 1 , then the clusters are {1,2},{3,4},{9},{8,7}.The clusters can also be {1,2},{3,4},{8,9},7.Any solution to these problem ??please reply . thanks .
Hello Madam, The two clusters can be formed with threshold 1. During implementation in the real world problem with clustering the other factors are also considered along with these factors.
Why do we call it "complete" and "single" linkage ? In both videos the difference was minimum and maximum distance to take after making the cluster, is there any other logical reason behind that naming ?
mam, everywhere it is written that the space complexity of naive Hierarchical Complete Linkage clustering algorithm is O(n) but as far as I know that if all the pairwise distances are stored for calculating the distance matrix then the space complexity should be O(n2).Will you plz let me know that why the space complexity is O(n)??
hi, why is it that we always have to find the minimum value in the lowest bound but at the last stage (11:53), just finding smallest value in the whole distance matrix?
![Hierarchical Agglomerative Clustering [HAC - Average Link]](http://i.ytimg.com/vi/T1ObCUpjq3o/mqdefault.jpg)
![Hierarchical Agglomerative Clustering [HAC - Average Link]](/img/tr.png)
![K-Mean Clustering [Single Dataset]](/img/n.gif)
You saved me, I got 37/40 in data mining. Thank you❤
Congratulations ❤️
Thank you for this walkthrough! It is very well done. I was looking every where to find an example of how to update the cluster matrix and this really helps. It is so well done and is extremely clear. Thank you for this series on clustering.
Thank you Sir.
exactly!!
Bro you look like Craig from Bearded Mechanic
Finally I understand the approach about how to merge these points.
You are saving my exam in data mining. Thank you very much!
God bless you for this simple yet informatic explanation of Agglomerative Clustering
SHe explained the Hierarchical Agglomerative Clustering very well. Big Thank
Hi at 3.50 , the euclidean distance should have 'y-b' instead of 'x-b' for the second value. Thanks and Nice explanation.
Yes Sir,
Thanks
Anuradha Bhatia Thank you for your teaching and it helps students like me in easily understandable steps.
Thank you so very much for motivation.
good lecture, a small mistake at 4:05 the euclidean distance formula should be sqrt(sum(x - a)^2 + (y - b)^2)
smart!
Best explanation of complete linkage I have found. Thank you so much!!!!
Thank you. I've been studying from a manual and this method not not even close to the explanation
Very clear explanation ma'am....and ma'am one big request please continue to do by centroid method....
It's very helpful for our ongoing exams....thank you so much ma'am.
Thanks.
My all doubts are clear now thx you so much. :)
Thank you very much for taking the time to post these really helpful videos!
I would like to thank you for this video. Your explanation is magnificent and so clearly. You helped me a lot to comprehend these complex subjects. Greetings from Mexico.
Thank you so much.
i like this video, good explanation, good step by step guide, i say more effective than what my teacher taught, thumbs up :3
SAME HERE
Hello, Thanks a lot for the simple and clear explanation for the single linkage (previous video) and for the complete linkage as well. I have two questions. 1) Looking at the dendrograms obtained from the single linkage and the complete linkage, one can see that they are different. So , how can we interpret that? If I cut the tree at the same value (i mean for the single and complete linkage trees), I would obtain different clusters results. 2) What is the idea behind searching for the maximum distance in case of complete linkage?
Thats for complete linkage.
Very lucid explanation. Keep up the great work !
Thanks Mam for explaining this. Very useful.
Thank you madam, very convincing explanation!
this was very helpful.. thank you
Thank you Maam. That was a clean video and helped me a lot understanding the Complete-Link. I have a few questions question..
1. How does the merge criterion influence the merge decision?
2. Why is this complete link clustering called non-local while the single link criterion called local?
Very good explanations! Can you please Show an example how to Use the correlation matrix as a distance matrix in kmeans. You have applied the euclidian distance in k means to cluster. How does the calculation of the Clusters work, with taking not the original dataset, but having the correlation matrix. How to Use the corr Matrix to Bild k means clusters? Thank you!
Thanks a ton for a fantastic explanation madam!
When we pick to first start the merging process, shouldn't we pick P6 and P5 to merge first since it has Max value 0.39?
The first cluster is determined through the most similar units. After that we define the distance from that cluster to the other data points through either single linkage (looking at minimal distance) or complete link (looking at maximum distance)
COMPLETE LINK - it means, while calculating distance matrix, we take the maximum value, right?
SINGLE LINK - while calculating, distance matrix, we take the minimum value?
Yes.
BEST OF LUCK.
Thanks :) means a lot, madam.
Pretty useful video! Can you share this slide?
Superb explaination .
does the formular at 3:44 contain a typo? should it be (y-b)^2? not (x-b)^2?
Thank you ma'am. It helps a lot
really helped! easy to understand the concept! thanks~
very very clear, thank you!
clean and precise video. Really helped. Thank you
Deepak Patter Thank you Sir
Deepak Patter 😊
why did we start with p3 and p6 ?shouldn't we start with the pair which has max distance between them?
Thank you for the amazing explanation.
What happens when you update your distance matrix and then there are two (or more) minimum values?
Thanks mam. Very well explained!!
Can you please explain what to do when the matrix has two same low value (eg: If P2 and P1 has 0.12 and P3 and P4 has 012). In that case which points need to be considered?
thank you so much for this, it really helped me!
Quick Question. How do we merge when the index of min element is 1,0 or 0,1
Thank you very much
Helped me a lot. Thank you.
mam i have a question: should we consider the least value or should consider the least value from the lower bound of the distance matrix
Try to upload classification sums naive Bayes ,bayessian and id3
ur video is very help full mumbai university students. ..try to solve it in same method followed by them .....Thanks
Thanks.
Sure Sir.
Thanks,clear explanation.
what to do if there are two same smallest elements?
Salam. Thanks a lot :). Excellent job.
dear mam, i would like to know that sometimes cases appear where after computing the similarity matrix we find two lowest distances . Now we can choose anyone of the distances to merge at that step. Now this decision may affect the cluster output at the final stage . Well here am talking about the case when a distance threshold is applied . say for eg-{1,2,3,4,9,8,7}.here if we take a threshold of 1 , then the clusters are {1,2},{3,4},{9},{8,7}.The clusters can also be {1,2},{3,4},{8,9},7.Any solution to these problem ??please reply . thanks .
Hello Madam,
The two clusters can be formed with threshold 1. During implementation in the real world problem with clustering the other factors are also considered along with these factors.
ok, so both of them can be the answers .. am i right?? now if any other factors are taken into consideration then we have to choose a single one..
Right.
Why do we call it "complete" and "single" linkage ? In both videos the difference was minimum and maximum distance to take after making the cluster, is there any other logical reason behind that naming ?
what if there is more than one smallest value? example : both value of (1,4) and (2,5) is 1
dear mam,
if we given a similarity matrix instead of distance matrix then what will be the approach?
regards
Atul
Amazing... Cheers
Thank so much...
Lifesaver!! Thank you :)
The answer or the final dendogram of both complete and average link can be same ?????
mahir khan Yes in few cases
Anuradha Bhatia thank you
3:15 - Complete Linkage
Awesome
mam, everywhere it is written that the space complexity of naive Hierarchical Complete Linkage
clustering algorithm is O(n) but as far as I know that if all the pairwise distances are stored for calculating the distance matrix then the space complexity should be O(n2).Will you plz let me know that why the space complexity is O(n)??
As every distance is computed and used exactly once.
would you please clarify slightly ..
hi, why is it that we always have to find the minimum value in the lowest bound but at the last stage (11:53), just finding smallest value in the whole distance matrix?
farah adilah lower bound...so smallest
oh.. so no matter complete/single/average linkage, always take the smallest value?
can anyone please share me the code for hierachical clustering
helps alot
Thank you!
Thank you madam
Thank you for the clear explanation :)
Thank you
thanks
well done .. thank u :D
Thanks! =D
Thanks!
❤️
y-b
Thank you !