👏👏👏👏 well done. Great explanation of the big picture and the relationship among statistical learning, probabilistic learning and the modern machine learning.
No. False positive means wrongly saying a patient has covid when he/she is not. Where as false negative means wrongly saying a patient does not have covid when he/she has.
@@Kandyfrom1986 Exactly. However, there seems to be a discrepancy with mathematical notation. TPR = P(T|C) and FNR = P(¬T|C) should add up to 1 (if a person has covid, then the test is either negative or positive), so given a TPR of 0.938, FNR should be 1 - 0.938 = 0.062. Similarly, TNR and FPR should add up to 1. Since here TNR = P(¬T|¬C) = 0.96, we can deduce that FPR should be 0.04. Thus, the task should say a "false positive rate". Please correct me if I'm wrong.
Given the card has one red face means we now have 2 possibilities So the probability that the other side is red should be 1/2 Please correct me if I am wrong.
That is incorrect, because you could have drawn card 1 on either side, so there is actually 3 possibilities. You have three cards and six sides: 1 2 3 top side R W W bottom side R R W When you drew a card with a red side you know that card 3 no longer plays a role in the probabilities. So there is three possibilities left: Ýou drew either : - the bottom side of card 1 in which case the other side is red. - the top side of card 1 in which case the other side is again red. - the bottom side of card 2 in which case the other side is white. This gives you a probability of 2/3 for the other side being red.
I’m thinking he is right. With one card already removed, two cards remain. One would be red and the other is white. That’s 50% because only two choices remain. However you want to talk around or over this doesn’t change a thing. The prof is wrong. There are not 3 possibilities. The red card is red on both sides. The white is white on both sides. If you want to be ridiculous you can say since there are 4 possible sides which can emerge next, and 2 are red and 2 are white, making 2/4 fraction and it’s 50% again. Once one card is first pulled there are no longer 3 choices.
I read somewhere there are JAX assignments for this course. Are they available to the public? Without solution is fine. These would help us reinforce the lecture material.
Would it be possible to also upload the codes from the lecture? I only saw the applet code but not the jupyter notebook codes. It would really help if the codes are uploaded too.
It would've been good not to say "upper side" and stress on the fact that no distinction can be made between faces. Cause if the sides can be distinguished, then the answer is 1/2.
If the sides are distinguishable, then 1/3 of the time you will be certain that you've got red-red card (since red-white card has this side white), and the remaining 2/3 of the time you will have indeed 1/2 probability you've got red-red card. Thus the final probability is: 1/3*1 + 2/3*1/2, which, funny enough, equals 2/3 :)
At 44:00 why does the professor say this works for any non zero probability set? I feel P(A ∪ Φ) = P(A) + P(Φ) and thus P(A) = P(A) + P(Φ) and so P(Φ) = 0 for any P(A)
![Seungmin "그렇게, 천천히, 우리(As we are)" | [Stray Kids : SKZ-PLAYER]](http://i.ytimg.com/vi/kAzmhLHePqU/mqdefault.jpg)
27:20 Really impressive, especially the pronunciation of Kolmogorov's name.
Great lecture and great lecturer. Thank you for making this available to public for free.
👏👏👏👏 well done. Great explanation of the big picture and the relationship among statistical learning, probabilistic learning and the modern machine learning.
These are great lectures, thank you so much. Are the latest lectures for Statistical Learning Theory also available on RUclips?
I am also quite interested on that ... I only found the 2020 version ruclips.net/p/PL05umP7R6ij2XCvrRzLokX6EoHWaGA2cC
Shouldn't it say "false positive rate" in the Covid19 task specification at 18:30?
No. False positive means wrongly saying a patient has covid when he/she is not. Where as false negative means wrongly saying a patient does not have covid when he/she has.
@@Kandyfrom1986 Exactly. However, there seems to be a discrepancy with mathematical notation. TPR = P(T|C) and FNR = P(¬T|C) should add up to 1 (if a person has covid, then the test is either negative or positive), so given a TPR of 0.938, FNR should be 1 - 0.938 = 0.062. Similarly, TNR and FPR should add up to 1. Since here TNR = P(¬T|¬C) = 0.96, we can deduce that FPR should be 0.04. Thus, the task should say a "false positive rate". Please correct me if I'm wrong.
@@iamkzntsvMy bad. You are right. Thanks for the correction
yep, should be "false positive rate" of 4%
Given the card has one red face means we now have 2 possibilities
So the probability that the other side is red should be 1/2
Please correct me if I am wrong.
you are wrong, and you were corrected in the lecture if you were actually listening to it.
That is incorrect, because you could have drawn card 1 on either side, so there is actually 3 possibilities. You have three cards and six sides:
1 2 3
top side R W W
bottom side R R W
When you drew a card with a red side you know that card 3 no longer plays a role in the probabilities. So there is three possibilities left:
Ýou drew either :
- the bottom side of card 1 in which case the other side is red.
- the top side of card 1 in which case the other side is again red.
- the bottom side of card 2 in which case the other side is white.
This gives you a probability of 2/3 for the other side being red.
I’m thinking he is right. With one card already removed, two cards remain. One would be red and the other is white. That’s 50% because only two choices remain. However you want to talk around or over this doesn’t change a thing. The prof is wrong.
There are not 3 possibilities. The red card is red on both sides. The white is white on both sides. If you want to be ridiculous you can say since there are 4 possible sides which can emerge next, and 2 are red and 2 are white, making 2/4 fraction and it’s 50% again.
Once one card is first pulled there are no longer 3 choices.
I read somewhere there are JAX assignments for this course. Are they available to the public? Without solution is fine. These would help us reinforce the lecture material.
can we have the homeward or exercises of this course?
Would it be possible to also upload the codes from the lecture?
I only saw the applet code but not the jupyter notebook codes. It would really help if the codes are uploaded too.
Can we get course website
Hi , Please provide code and excercise of this very nice code.
It would've been good not to say "upper side" and stress on the fact that no distinction can be made between faces. Cause if the sides can be distinguished, then the answer is 1/2.
If the sides are distinguishable, then 1/3 of the time you will be certain that you've got red-red card (since red-white card has this side white), and the remaining 2/3 of the time you will have indeed 1/2 probability you've got red-red card. Thus the final probability is: 1/3*1 + 2/3*1/2, which, funny enough, equals 2/3 :)
Are there plans to upload these slides? The previous ones are from 2020.
The material is now avaiable at github.com/philipphennig/Probabilistic_ML
@ Is the Statistical Learning Theory lectures and slides going to be uploaded as well?
So sick
At 44:00 why does the professor say this works for any non zero probability set? I feel P(A ∪ Φ) = P(A) + P(Φ) and thus P(A) = P(A) + P(Φ) and so P(Φ) = 0 for any P(A)
man's mind ❌ Human mind ✅