Don't know why most profs spend so much time no theory and almost none on examples. This really hammered the concept in for me. Excellent videos, keep it up!
@@danielvazquezguevara3842 I think in conditional probability and Bayes nets, there are active paths, rather than nodes. An active path between 2 nodes indicates they are NOT conditionally independent of each other.
This is the best learning video I have ever watched. The many samples and repetition are what students really need in order to understand a concept. Extraordinary work! Subscribed
All hail Pieter Abbeel for explain the otherwise inexplicable 🙏🙏🙏 I really appreciate his patience to explain the details for each example; exactly what I needed!
Thank you very much for this great online lecture Professor Abbeel. Finding this video earlier would have saved me a lot of time and headache trying to decypher cryptic university scripts :D
Why my university can't simply put the link to this video in the lecture slides and not bother the teachers who can't really teach? Thank you Prof. Abbeel!
It only takes 1 discovered active path to make independence not guaranteed. So even if every other path they checked was inactive, that first active path invalidates any guarantee of independence
can someone help me with the intuition behind classifying the active and the inactive triples? For eg: What makes the causal link to be inactive if the middle node is observed?
Very nice to have so many examples! That said, it would perhaps have been easier to comprehend if different colors were used to distinguish paths from triples. The paths seem to get muddled together with the triples and the triples are hard to see or distinguish for some of the examples. Thanks for the fantastic lecture!
can you please explain what do you mean when you're saying in example 2 "T is observed". i can't understand what do you mean by saying observed. thanks in advance
Observed node means you know the value of the random variable affected to this node. For instance, if the node is "It's raining" (True/False), and if in your situation you know that yes, it's raining, then the "It's raining" node is observed. Conversely, if you know that it's not raining, the node is also observed. If you don't know whether it's raining or not, then the node is NOT observed. Think of a node as a sensor which can take two or more different values. If you know this sensor's value, it means that you can "read" it. In other words, this sensor is "observed" by you (otherwise you couldn't know).
Yeah this video has a number of consistency problems, and doesn't go into detail about the difference between each example. Below are some errors I found, either they are errors in what is explained (contradicts the logic being used) or they are errors in the solution (the logic is wrong). At 8:47 there is an error. Why would U _||_V | Y be "Active" when it contains the V-structure W->YYYY
Even if there is only one active path among all possible paths between two nodes than you can say that independence is not guaranteed to be true between these nodes. At 8:47 U _||_V | Y Y is observed and it makes path U-W-V active, so you don't need to consider any other paths and say independence is not guaranteed to be true. At 10:05 there are two paths, one of them is WX, lets consider this path. There is the only triple on this path, W-V-X and it is active. It means you don't need to check the second path and say that independence is not guaranteed to be true.
At 8:47 there is not an error. We found an active path between U & V and thus we can draw the conclusion that we can not guarantee independence. The v structure's inactivity is irrelevant since we have already found an active path between U & V. At 6:58 the quotation you have written doesn't make any sense and is not said in the video? At 6:05 your quotation is completly wrong. He says that an inactive TRIPLE implies and inactive PATH. Which is what he has been saying in nearly every example. If there is an inactive triple and an active triple in a path then the path is inactive for your clarification. At 10:05 he says that it is the ONLY TRIPLE IN THIS PATH not that it is the only triple. He has already explained why he ignores the bottom path. It is because he found that the first path is active so he doesn't have to check the other paths. Because, if at least one path is active then we can not guarantee independence. The logic with the triples is clearly explained in the beginning of the video .Maybe you didn't pay attention in the beginning? My recommendation to you would be to try to rewatch the beginning and try to remember the triples that are inactive and active as well as trying to understand the difference between a triple and a path. Best /T
This is great. Glad there were so many examples to really drill the concept home. Exactly what I needed.
same!
Don't know why most profs spend so much time no theory and almost none on examples. This really hammered the concept in for me. Excellent videos, keep it up!
Can’t thank you enough. This is the only video people need to learn about d-separation. Thanks again for your time and effort!
Really awesome video! Numerous examples made learning much easier rather than just listening to lectures on the concepts.
this is the best video on the internet that explains d-separation!! Thanks a lot Peter!
by chance do you know what does it mean to be active node or inactive node?
@@danielvazquezguevara3842 I think in conditional probability and Bayes nets, there are active paths, rather than nodes. An active path between 2 nodes indicates they are NOT conditionally independent of each other.
@@viveksmenon123 Thank you!
This is the best learning video I have ever watched. The many samples and repetition are what students really need in order to understand a concept. Extraordinary work! Subscribed
Its been more than 10 years now when this video was uploaded. Still I find this the best video with so many examples! great.. Thanks!
All hail Pieter Abbeel for explain the otherwise inexplicable 🙏🙏🙏 I really appreciate his patience to explain the details for each example; exactly what I needed!
This is the first video in which I now totally get it. Thank you so much 😊
:)
I'd like to join with the other folks that have expressed you their gratitude for this video, finally I get it.
Wish you the best.
Amazing man!! The best video to really understand d-seperataion!! and only in 20 minutes!! Amazing
Thank you very much for this great online lecture Professor Abbeel. Finding this video earlier would have saved me a lot of time and headache trying to decypher cryptic university scripts :D
Bless you sir for your many step-by-step examples. I could not understand D-Separation until I saw your video!
You are Awesome. You are godfather of machine learning
Amazing, I was struggling with this topic, this video made me 100% clear
Before watching your video, I found it hard to apply D-separation. But thanks to your video, everything is crystal clear now.
This is the best tutorial I've ever seen, thank you!
Thank you so much for this video! I’ve passed my last exam only thanks to you!
Such a great resource! Thank you for your time to prepare this.
Glad it was helpful!
This is awesome, thanks for making this video lot of clarity gained and cleared after watching this.
Why my university can't simply put the link to this video in the lecture slides and not bother the teachers who can't really teach? Thank you Prof. Abbeel!
Thanks dude! Now I got it! They have never told me about this extended collider rule.
Thank you very much. Youre really helping me learning for my exam thursday :)
All the examples make it easy to understand.
This was very helpful and well explained. Thanks for taking the time to make this lecture!
you are a hero.. saved my exam
You sped through example at 7:00 really quick, only checking the u->w
It only takes 1 discovered active path to make independence not guaranteed. So even if every other path they checked was inactive, that first active path invalidates any guarantee of independence
Thanks! This was an excellent video!!
can someone help me with the intuition behind classifying the active and the inactive triples? For eg: What makes the causal link to be inactive if the middle node is observed?
This is exactly what all students need ! awesome ! great thanks :-)
This video was insanely helpful, THANK YOUUUUU!!!
The video was very intutive thanks a lot for explanation !
Thank you sooo much sir. God bless you. You saved me.
This is a very nice video to learn the two rules of de-separation
Thanks for all your examples
Thank you for this video. It really helped me to understand the topic
Very nice video! keep up the work
Thank you very much!
You saved me from failing my exam !!
This is such a wonderful video!!!.. Thank you so much 🙏
Thanks for the awesome, elaborated explanation!
This video helps me as well as my friends.
Thanks a lot :D
thank you thank you for this vedio. after watching this vedio i really get the point of d-seperation~ : )
very good one. I grasp the concept now.
You saved my life
Love the examples, am currently prepping for an AI midterm
Awesome! Thank you! Finally understood the topic !
Very nice to have so many examples! That said, it would perhaps have been easier to comprehend if different colors were used to distinguish paths from triples. The paths seem to get muddled together with the triples and the triples are hard to see or distinguish for some of the examples. Thanks for the fantastic lecture!
This video is really amazing. Thank you!
Great video. Simple and on point.
great video, thank you very much!
thanks Pieter! Amazing explanation. from Matteo and Lilli
Great video! Very helpful! Thank you!
wow so many examples thanks!
thank you for great explanation.
Could you elaborate 8:54? is it really active?
Comprehensive and understandable. It's Great.
Just what I was looking for, thank you so much
1:20, what do you mean by "observed"?
Thanks for the help!
Sad I can only give 1 like because this video was A M A Z I N G !
I literally understood in 5min
Thank you for explanation.
thank for your video. But i have a question. How to know which variables can be observed or unobserved ?
Gosh!!! This video literally saved my ass
This is awesome!! Thanks Prof.
can you please explain what do you mean when you're saying in example 2 "T is observed". i can't understand what do you mean by saying observed. thanks in advance
What's the difference between an observed and an unobserved node?
Observed node means you know the value of the random variable affected to this node. For instance, if the node is "It's raining" (True/False), and if in your situation you know that yes, it's raining, then the "It's raining" node is observed. Conversely, if you know that it's not raining, the node is also observed. If you don't know whether it's raining or not, then the node is NOT observed.
Think of a node as a sensor which can take two or more different values. If you know this sensor's value, it means that you can "read" it. In other words, this sensor is "observed" by you (otherwise you couldn't know).
You are legend!
thankuuuuuuuu so much for great explanation
This video is excellent. Kudos!
Wonderful wonderful video.
It helped me a lot! Thanks
So helpful, thank you very much
Excellent. Thank you so much!
A necessary clarification I found in a comment here:
Even an inactive triple makes a path (that might have multiple triples) inactive
thanks that helps a lot
Good explanation
This is great! Thanks!
Thank you for saving my ass, great video.
What if the nodes are adjacent?
If you can't understand after so many examples then just leave it man find something else to study
@@ane-ct9xy you are probably right.
So clear... Thank you!
Very helpful! Thank you!
thank you very helpful to me
I wish you were teaching cs188 again!
Awesome !! Thank you so much :)
thank you
Very clear. Thanks.
Very helpful , Thank you
great. tahnks
Thank you!
🙏
Useful, thank you.
Great
🤩
yea i have a test 8 days from now, thank you
Yeah this video has a number of consistency problems, and doesn't go into detail about the difference between each example. Below are some errors I found, either they are errors in what is explained (contradicts the logic being used) or they are errors in the solution (the logic is wrong).
At 8:47 there is an error. Why would U _||_V | Y be "Active" when it contains the V-structure W->YYYY
Even if there is only one active path among all possible paths between two nodes than you can say that independence is not guaranteed to be true between these nodes. At 8:47 U _||_V | Y Y is observed and it makes path U-W-V active, so you don't need to consider any other paths and say independence is not guaranteed to be true. At 10:05 there are two paths, one of them is WX, lets consider this path. There is the only triple on this path, W-V-X and it is active. It means you don't need to check the second path and say that independence is not guaranteed to be true.
At 8:47 there is not an error. We found an active path between U & V and thus we can draw the conclusion that we can not guarantee independence. The v structure's inactivity is irrelevant since we have already found an active path between U & V.
At 6:58 the quotation you have written doesn't make any sense and is not said in the video?
At 6:05 your quotation is completly wrong. He says that an inactive TRIPLE implies and inactive PATH. Which is what he has been saying in nearly every example. If there is an inactive triple and an active triple in a path then the path is inactive for your clarification.
At 10:05 he says that it is the ONLY TRIPLE IN THIS PATH not that it is the only triple. He has already explained why he ignores the bottom path. It is because he found that the first path is active so he doesn't have to check the other paths. Because, if at least one path is active then we can not guarantee independence.
The logic with the triples is clearly explained in the beginning of the video .Maybe you didn't pay attention in the beginning? My recommendation to you would be to try to rewatch the beginning and try to remember the triples that are inactive and active as well as trying to understand the difference between a triple and a path.
Best
/T
Such a confusing video. Making it unnecessarily difficult to understand basic concepts.