@GetMeThere1 Yes, to get *accurate* estimates of the filter parameters, programmers could only realize it if they use Bayes' Theorem. That is the 'math' part that the presenter skipped over. Notice that when he said the sample votes were 7 Yes and 3 No, he said the prob of the overall vote being Yes was 89%. But if you just use a naive calculation, you might incorrectly estimate the prob as 7 / (7 + 3) = 7/10 = 70%. To get the correct value of 89%, you need Bayes' Theorem.
I agree w/ GetMeThere1. The idea of multiple or accumulative criteria for discerning something must have existed before probably with no label. If a stranger came to town on a type of wagon, with a certain kind of sword with a certain accent with one or more types of currency etc. c'mon.
Why irks me about something like this is that its totally OBVIOUS and common sensical. Do you mean to tell me that programmers could only REALIZE this if they had a TERM (Bayes Theorem) to work from?
So, if I were going to design a filter I'd say "Let's find a bunch of filterable parameters, and figure out how well it worked when 'x number' or 'y combination' of parameters 'hit.'"
And I wouldn't have to know ANYTHING about 18th century ministers.
The election example is bad, it's not Bayes'.
walawala147 1 month ago
i wonder if the election example is related to bayes' theorem?
it looks like estimating population proportions with sample proportions using normal distributions, which i learned to be a different issue
fazekaslaszlo 1 year ago
RIP Mr. Artner
ecomsys1984 1 year ago
@GetMeThere1 Yes, to get *accurate* estimates of the filter parameters, programmers could only realize it if they use Bayes' Theorem. That is the 'math' part that the presenter skipped over. Notice that when he said the sample votes were 7 Yes and 3 No, he said the prob of the overall vote being Yes was 89%. But if you just use a naive calculation, you might incorrectly estimate the prob as 7 / (7 + 3) = 7/10 = 70%. To get the correct value of 89%, you need Bayes' Theorem.
wonderist 1 year ago
Thanks for introducing a practical application of Bayes' Theorem!
PQK 2 years ago
I agree w/ GetMeThere1. The idea of multiple or accumulative criteria for discerning something must have existed before probably with no label. If a stranger came to town on a type of wagon, with a certain kind of sword with a certain accent with one or more types of currency etc. c'mon.
trombone7 2 years ago
it's not really statistics - it's probability
thelastwords 2 years ago
haha this is neither. its actually statistics
fakeaccount004 2 years ago
is this algebra or calculas?
hotelroyale100 2 years ago
Why irks me about something like this is that its totally OBVIOUS and common sensical. Do you mean to tell me that programmers could only REALIZE this if they had a TERM (Bayes Theorem) to work from?
So, if I were going to design a filter I'd say "Let's find a bunch of filterable parameters, and figure out how well it worked when 'x number' or 'y combination' of parameters 'hit.'"
And I wouldn't have to know ANYTHING about 18th century ministers.
GetMeThere1 3 years ago