The weirdest paradox in statistics (and machine learning)

Video quality	The size	Download

Information The weirdest paradox in statistics (and machine learning)

Title	:	The weirdest paradox in statistics (and machine learning)
Lasting	:	21.44
Date of publication	:
Views	:	1 jt

Thanks!
Comment from : @sgtravz

I have to do a real world example using a Stein’s estimator for a parallel experiment, does someone has any idea or help? brThank you in advance
Comment from : @giovannimantovani795

I'm taking advanced econometrics this semester and this James-Stein approach came up as one of the example of shrinkage estimator Felt like spotting an easter egg in a movie haha
Comment from : @jch123-r6v

Thank You
Comment from : @jakeaustria5445

This explains why what mice eat for lunch affects black holes 10,000,000,000 light years from Neptune
Comment from : @AspartameBoy

less mysterious perhaps than it seems in a certain ineffable sends in my mind
Comment from : @jpphoton

PI and PHI
Comment from : @PhiGuy01

Who could ever guess that shrinkage could be useful in higher dimensions?
Comment from : @Melorama2000

Fantastic!
Comment from : @vtrandal

Wow it is completely unreal how good this video is
Comment from : @tostupidforname

He said Mewtwo he said mewtwo
Comment from : @Domainexpansioncolon3

I see what you did there :) "Let's take two random points"brbPicks Pi and the golden ratio/b
Comment from : @NoName-md6fd

Ronald Fisher 人类高质量男性😃
Comment from : @macknightxu2199

Is this saying that by knowing something about any entirely separate and unrelated system you will know more about every pair of another two? That means they're related because each one provides info about the other two, which contradicts the premise We can infer there is no entirely independent system Is that really what's happening?
Comment from : @tricky778

I am typing this without using the brain because my brain exploded and is now flowing down by the wall
Comment from : @Sergiuss555

Your videos are far more confusing than your competition 😅
Comment from : @teckyify

The best estimator is infinity br-Calculus
Comment from : @4doorsmoorhoors542

Whrs Yp-stein Im-stein confused
Comment from : @godalmighty5970

Do woman know about the shrinkage?
Comment from : @MateBabic

I like this video very much but for me it seems unintuitive to throw different independent problems together and evaluating them as one
Comment from : @사이보그-i6p

Mental masturbation makes you go blind and makes your tool shrink
Comment from : @MyNomDePlume

I wish I fully understood Fun to see the maths!
Comment from : @robert-wr9xt

Static 0:59 bcaa CBRN apex shocks 2:09 WIMP filtered politics 3:00 variable quality vs survival evolving cordons 6:21
Comment from : @thaq82

Excellent content, research, pace and presentation Thank you for putting this together and explaining it in simpler terms than the paper :)
Comment from : @GeorgeZoto

Best estimator for dataset mean is summ of data point values divided to number of data point Paradox solved
Comment from : @MikeRebelScum

I love unintuitive concepts in all branches in mathematics Its actually the motivating reason why I studied mathematics in undergrad!
Comment from : @TheMrlightswarm

Currently studying ML and went over Thompson Sampling recently This is a great video as i immediately saw the similarities and was able to follow along even though i knew nothing about ML before i got started Definitely subscribing
Comment from : @kel3747

The one thing that makes statistic's useless is personal biasbrEven when applied to algorithm and poorly coded AI like Chatgpt , when code is written bye individual biological units , biases are inherent in the software, Like chatGpt being a racist piece of white software
Comment from : @ossiedunstan4419

How to avoid hard work and use useless data by fetishizing and obfuscating one's profession
Comment from : @polkabike

GENUINE IDEAS
Comment from : @gyanprakashraj4062

I'm at 3:35 strangely I am unsurprised My logic so far is that I have three guesses I imagine that all three guesses have an error, I reckon some are more errorish than others, so I flatten the risk over all three data points by @averaging the risk@ I shall continue the video to see If I get surprised later on Certainly I am interested to find out if my concept is remotely like theirs
Comment from : @redfernpixelgnomepitcher1377

you copied this from 3blue1brown
Comment from : @yessure5792

Mewtwo, I choose you!
Comment from : @thebluriam

I stumbled into this and I still dont know wtf I just watched
Comment from : @SpiritGear

It would be cool if the terms used in this video were explained a bit Up until 13:00 I thought that mu is meant to be value at maximum, not the point at which maximum occurs, which was pretty confusing
Comment from : @pawedziedzic3250

dont watch
Comment from : @stewiegriffin6503

Omg, from now on this is how I am gonna explain regularization and bias-variance trade off to my students
Comment from : @AbhishekSingh-gk9gh

have to finish to watch video, on up to 2:17 it's annoying that two "randomly chose points" somehow yielded Pi and Pi/2, what are the chances of that?
Comment from : @hams3r_dont_upload_anythin182

I have herpes
Comment from : @handsome_man69

if you have only one data point you do not estimate µ, instead you get a second data point
Comment from : @DasJev

@2:55 Are those values even the same order of magnitude??
Comment from : @pj_ytmt-123

The value of mu could not be on the left of 314, because you established it at the beginning to be at the top of the bell curve
Comment from : @chaoticignorant483

3:13 the way you phrased this makes it sound like i can better estimate whether my crush likes me if i also take under consideration the alignment of the planets on the day of her birth and how much wood a given woodchuck chucked, given that a woodchuck would chuck wood
Comment from : @Navak_

Why 314? Can someone explain?
Comment from : @ktsai2003

If I had one data point that was pi and another that was phi, I would probably start believing in God
Comment from : @Phi1618033

It does not always result in shrinkage right? When x are generally small enough so that the shrinking factor you have could be smaller than -1 which would result in a enlargement instead of shrinkage In that case it’s no longer better than the maximum likelihood estimator
Comment from : @廖浩宇-m2r

I'm confused Are you a jew living in Holland with a speech impediment? Is incredibly hard to listen to you speak It's like you are guessing at how to pronounce every word
Comment from : @Tadesan

I'm sorry if I'm just stupid but even after watching the entire video I still don't see how 3 random sets of data can have any impact on the estimation of any other set of data That's like saying if the grass is a random color than the sun must be red It doesn't make any sense practically Unrelated information doesn't just suddenly relate to each other
Comment from : @KingWesleyIV

This video is colddddd 🔥 🔥 🔥
Comment from : @MedicalOkami

Ool
Comment from : @NuEu-er5wf

???????????????????
Comment from : @jackdaniels5538

😮
Comment from : @ilyboc

All this video is like "it works because it works"
Comment from : @okolenmi7511

So this is why I continue to hate the way statistics is taught - despite it actually being indispensable What a terrible explanation
Comment from : @mjcau

Great video but bvery hard to make out what you’re saying/b Get speech therapy/coaching and your videos will get millions of views
Comment from : @richardhouseplantagenet6004

I don’t understand but ok
Comment from : @bigblackbadger1

I got into a fist fight with my wife about this very thing the other day She won the fight, but in the end, I was right
Comment from : @jaykemm3472

you should put more ads I am clearly able to follow the video and not distracted enough I hit the bloody monitor just 3 times You must drive me to throwing my pc out of the window You can do it Just put ads every 40 seconds instead of every 2 minutes
Comment from : @sergche3718

I somehow feel this is really connected to the concept of higher dimensional spheres, which 3b1b hadmade a video on About their volumes and shapes
Comment from : @chaitanyalodha3948

They are just taking the fact that if you have points missing left and points missing right, the average error will help decrease your total error
Comment from : @Trizzer89

I'm confused When you have 3 or more data sets does this help you predict the average of their mu taken all together Or just each one independently?
Comment from : @ChucksExotics

Steam engines on top of reactors and/or chimneys made from renewable resources like silicone nitritde can produce more energy on location and set a absolute zero emissions standard
Comment from : @josephhuelbig

If instead of containing running water in circuits they had a run off on full pressure into the ocean that they meditate as the need arises there would not be a a low oxygen level in the ocean
Comment from : @josephhuelbig

Dumping compost into the oil drilling holes can creat a crude oil substance that should potentially deter a global meltdown from the water that's been placed into the holes They misunderstood the oil-water sepation and evaporation for insulation
Comment from : @josephhuelbig

To fix the hole in the ozone you need a fresh water cloud and ice seeding process on a pattern that follows the hole Dropping ice into the hole from the "top" and seeding clouds in the upper, middle and lower accumuls of clouds
Comment from : @josephhuelbig

You're a liar
Comment from : @lordsneed9418

Now I finally understand L1 and L2 Regularisation
Comment from : @prakharkumar4114

Missed opportunity to coin a new term, "bestimator"
Comment from : @frozencreed

I have absolutely negative zero idea what you are talking about
Comment from : @miketacos9034

haha mew2
Comment from : @daysound9761

You did a poor job explaining this after 5:00you should have used many more examples and related it to concrete ideasfor example, you gave a poor explanation of mean square error and why we were plotting anythingyou sound too much like a textbookI've noticed this on many of your videos Sorry, you're just not a good teacher
Comment from : @kidzbop38isstraightfire92