I am preparing my data mining exam for wednesday and this helped me understanding how the heatmap can be used. The video is incredibly clear, thank you so much!
Honestly I didn't find even a decent enough explanation of clustermap elswhere. This saved me. Infact I realized that I had understood dendrograms wrongly. I could see quite a lot of technical detail in your videos.
Excellent! Yes, I didn't find much explanation out there about the underlying hierarchical agglomerative clustering either - likely because seaborn is inheriting that from scipy and fastcluster. Glad to hear my explanation was useful!
Thank you soo much for making this video! I just had a question regarding what the default linkage method is for seaborn. I was also wondering whether we can numerically calculate the distance between columns or rows.
Useful and straight to the point, thanks! To your knowledge is there a way to have an OnHover functionality to display the x,y label? With many items it becomes hard to use the axis labels.
I didnt know that one exists, looks pretty interesting, thank you very much! I followed your tutorials for quite a while and recently i saw some videos about altair. The simple syntax reminded me about seaborn and it seems to be even more simple because you can also aggregate within altair and not use pandas before like you would do in seaborn. I also liked the hole idea behind it with the grammer for the visualisation. I curious what you think about altair?
Awesome - thanks for following along! Yes, I have heard some people say they like Altair, and one of my students used it recently. I haven't used it much personally, but the code I've seen looks fairly straightforward. I will definitely have to check it out and maybe make some videos about it someday! 😄
Sure thing - there are various distance metrics you can use to create the clusters, but the most common is Euclidean distance. That's just your regular distance from point A to point B that follows the typical distance formula. In the seaborn clustermap, that's the default distance (metric='euclidean'). So to follow the hierarchy we just merge together the two closest clusters. The tricky part is that sometimes the clusters contain multiple points, so how do you measure distance between clusters? There are several different methods. You could measure the distance from one cluster to the other in terms of the closest points... or the farthest ones away... or perhaps based on the average distances. All of these options are available via seaborn clustermap's "method" argument. The default method is by averaging the points within each cluster. Cheers!
I'm gonna use this a lot in my Classification related notebooks. Thanks for the lucid and in-depth explaination of this plot. Keep going :)
Excellent! Glad you found it useful 😀
I am preparing my data mining exam for wednesday and this helped me understanding how the heatmap can be used. The video is incredibly clear, thank you so much!
Thank You, You are the Best in Seaborn!
Very clear super understandable and in-depth
Great to hear! Glad it was helpful.
Excellent explanations! Thank you!
Honestly I didn't find even a decent enough explanation of clustermap elswhere. This saved me. Infact I realized that I had understood dendrograms wrongly. I could see quite a lot of technical detail in your videos.
Excellent! Yes, I didn't find much explanation out there about the underlying hierarchical agglomerative clustering either - likely because seaborn is inheriting that from scipy and fastcluster. Glad to hear my explanation was useful!
Thank you for a great video with suggested clustered heatmap interpretation! Subscribed to your channel rightaway! :)
This was nice and simple to follow! do more do more! 😄
Thanks for this vídeo. Good explanation
Great tutorial! Thank you!
Most welcome - glad you liked it!
Thank you soo much for making this video! I just had a question regarding what the default linkage method is for seaborn. I was also wondering whether we can numerically calculate the distance between columns or rows.
Useful and straight to the point, thanks! To your knowledge is there a way to have an OnHover functionality to display the x,y label? With many items it becomes hard to use the axis labels.
отличная подача материала.
Hi Kimberly, great video. What would you suggest if there are some missing values (NA) in my matrix and I don't want to drop any samples out?
you are professional keep the work you gonna be big CZcamsr
Thank you very much - I appreciate that! 😄
I didnt know that one exists, looks pretty interesting, thank you very much! I followed your tutorials for quite a while and recently i saw some videos about altair. The simple syntax reminded me about seaborn and it seems to be even more simple because you can also aggregate within altair and not use pandas before like you would do in seaborn. I also liked the hole idea behind it with the grammer for the visualisation. I curious what you think about altair?
Awesome - thanks for following along! Yes, I have heard some people say they like Altair, and one of my students used it recently. I haven't used it much personally, but the code I've seen looks fairly straightforward. I will definitely have to check it out and maybe make some videos about it someday! 😄
Great video! In the penguin example, for each penguin, can we say what are the most important features from the dendogram?
Cool video!
Thanks - glad you enjoyed!
hey, goodjob!
Awsome!
Thank you - cheers!
Thank you for this... Can you explain how to calculate the cluster diatances
Sure thing - there are various distance metrics you can use to create the clusters, but the most common is Euclidean distance. That's just your regular distance from point A to point B that follows the typical distance formula. In the seaborn clustermap, that's the default distance (metric='euclidean'). So to follow the hierarchy we just merge together the two closest clusters. The tricky part is that sometimes the clusters contain multiple points, so how do you measure distance between clusters? There are several different methods. You could measure the distance from one cluster to the other in terms of the closest points... or the farthest ones away... or perhaps based on the average distances. All of these options are available via seaborn clustermap's "method" argument. The default method is by averaging the points within each cluster. Cheers!
need your guidance on networkX
Oh fun! I haven't used networkX for a while, but the last time I used it, I enjoyed it. Will definitely consider making some videos about it! 👍
Great tutorial, thank you!
Thanks - cheers!