(ML 16.3) Expectation-Maximization (EM) algorithm

[@pemfiri]

I have seen many of your videos, i just wanted to say how much i appreciate the collections you have put up in here. they are very very useful and intuitive. thank you so much for all the time you have put forth to produce them.

[@rubenvanoostenbrugge6065]

My strategy: until_understanding{1. watch youtube videos 2. read literature}

[@oldcowbb]

it's converging very very slowly

[@po-yupaulchen166]

Thanks for your video. You totally understand the problem and have an excellent explanation. However, I have one comment about 'Why EM makes sense'. I think it might be more easily explained by Majorization-Minimization (MM) procedure. MM is an 'intuitive' optimization method, which approximates your cost function 'p_{\theta}(x)' in each iteration. And the approximation function is your Expectation procedure.

[@smoghe29]

I'm attempting how to apply EM algorithms to a data clustering problem. I understand (or at least seem to) the theory but struggle when converting the formulas to a practical problem. Can you help with any specific numeric examples ?

[@linkinpark2041]

Im quite confused about "taking the expectation value with respect to theta t" , while theta t is not a random variable. . should it be taking the expectation value with respect to Z?

[@adamnlee1886]

theta_t is an estimator, so it always is a random variable. The underlying logic is that any estimator depends on data, which itself is random. The real parameter theta, however, is not random.

[@allandogreat]

very good introduction for EM

[@misaprendizajes]

What does it mean the sumatory over z of p(x,z)? If Z is a random variable, how can you do a sumatory over de values of z? can somebody explain me it?

[@clairesmith9637]

It's the summation over all the potential values that z could take.

[@hadadvitor]

Super nice explanations. Thanks for all this.

[@statisticstime4734]

Excellent!

[@wavegeekman]

Totally lost me at the expression for the E step.

[@DioFerrari]

good video. it applies to non-exponential family as well

[@bigears8296]

How do you know data is missing apriori?

[@mikel5264]

Great video

[@kanchy100]

how do you know which initial guess to make?

[@Jacob-jc6hj]

It is random initially

[@eyeonthewinner]

thank you, this topic can be confusing..

[@yuxuanzhang1021]

what is the software used in the video?

[@asaskui]

please add some explanation about MM algorithm. because you always make it look easier, it would be helpful. thank you very much.

[@austinfeight3001]

"let me give some examples later" doesn't give examples.

[@shubhamvij6146]

Examples are given in the following videos... i.e. Gaussian mixture model videos

[@kaourtishis]

ummm, what is Z?

[@allandogreat]

Hidden parameters

[@Felixombc]

This video provides nothing you cant get from looking at the wiki page for EM. No theory, no useful intuition, and very little direction.

[@heiismail]

make a dummy version i am totally lost

[@vincenzo4259]

Thanks

[@thekolbaska]

You are my role model! Thank you!

[@garyzhai9540]

No idea all these jargons are as if we are all experts.

[@awandererscry]

what program are you using to write out the notes?

[@awandererscry]

found it on your youtube page. thanks~

[@TeleporterDown]

Check out his channel. He gives equipment used and prices.

[@yallowrosa]

unclear ...

[@awandererscry]

great tutorial btw :D

[@RaviTejaC]

thank you

[@swordinhands]

great voice

[@GeneralSeptem]

So slow, even at double speed...

[@themohameDkhalifa]

Poorly explained