Hidden Markov Models 12: the Baum-Welch algorithm

[@059812]

Stop searching, this is the best HMM series on youtube

[@science_electronique]

Sure I confirm

[@juliocardenas4485]

Is this the original channel for the series ?

[@djp3]

@@juliocardenas4485 yup

[@kevinigwe3143]

Thoroughly explained. The best series I have seen so far about HMM. Thanks

[@djp3]

Great to hear!

[@ligengxia3423]

I don't think anyone is gonna hit a dislike button on this series of video. Prof Patterson truly explained the abstract concept from an intuitive point of view. A million thanks Prof Patterson!

[@Ob.605]

You are definitely a life saviour! One can be studying about EM and HMM for a long while, but the need to go back to the basics is always there.

[@simonlizarazochaparro222]

I love you! I listened the lecture of my professor and I couldn't even understand what they were trying to say. I listened to you and things are so clear and easily understandable! I wish you were my professor! Also very entertaining!

[@djp3]

Glad I could help!

[@rishikmani]

whoa, what a thorough explanation. Finally I understood what Xi is! Thank you very much sir.

[@djp3]

Glad it was helpful! I wish I had pronounced it correctly.

[@benjaminbenjamin8834]

This is the best series on HMM, not only the Professor explains the concept and working of HMM but most importantly he teaches the core Mathematics of the HMM.

[@veronikatarasova1314]

Very interesting, and the examples and the repetitions made clear the topic I thought I would never understand. Thank you very much!

[@djp3]

You're very welcome!

[@idiotvoll21]

Best video I've seen so far covering this topic! Thank you!

[@djp3]

Glad it was helpful!

[@marlene5547]

You're a lifesaver in these dire times.

[@hannahalex3789]

One of the best videos on Baum-Welch!!

[@linkmaster959]

One of the main things that has always confused me with HMM's is the duration T. For some reason, I thought the duration T needed to be fixed, and every sequence needed to be the same duration. Now, I believe I finally understand the principles of the HMM. Thank you!

[@vaulttech]

There is a good chance that I am wrong, but I think that your description of Beta is backwards. You say (e.g., at 7:40 ) it answers "what is the probability that the robot is here knowing what is coming next", but it should be "what is the probability of what is coming next, knowing that I am here". (in any case, thanks a lot! I am trying to learn this in details, and I found the Rabiner paper quite hard to digest, so your videos are super helpful)

[@vineetkarya1393]

I completed the course today and it is still the best free material for learning hmm. Thankyou professor

[@djp3]

I'm glad it was helpful. This is a tough concept

[@garimadhanania1853]

best lecture series for HMM! Thanks a lot Prof!

[@ribbydibby1933]

Doesn't get much clearer than this, really easy to follow!

[@SStiveMD]

Astonishing explanation! Now I can resolve and understand better my homework for Knowledge Representation and Resoning

[@djp3]

Glad it was helpful!

[@sheepycultist]

My bioinformatics final is in two days and im completely lost, this series is helping a lot, thank you!

[@djp3]

Good luck. Hang in there! There's no such thing as "junk" DNA!

[@comalcoc5051]

Thanks proff really help me understand HMM on my research. Hope you have a good life

[@djp3]

Pay it forward!

[@barneyforza7335]

This video comes up so far down on the searches but is good (best) xx

[@Steramm802]

Excellent and very intuitive explanations, thanks a lot for this amazing Tutorials!

[@SPeeDKiLL45]

Thanks so much. Very talented in explaining complex things.

[@matasgumbinas5717]

There's a small mistake in the equation for the update of b_j(k), see 22:37. In both, the denominator and the numerator, gamma_t(i) should be gamma_t(j) instead. Other than that, this is a fantastic series!

[@djp3]

Yup you are right. THanks for the catch

[@leonhardeuler9028]

Thanks for the great Series. This series helped me to clearly understand the basics of HMMs. Hope you'll make more educative videos! Greets from Germany!

[@djp3]

Glad it was helpful!

[@IamUSER369]

Great video, thanks for clearing up the concepts

[@djp3]

My pleasure!

[@shabbirk]

Thank you very much for the wonderful series!

[@karannchew2534]

14:30 Why is bij (Ot+1) needed? aij = the probability of moving from state_i to state_j βt+1(j) = probability of being at state_j at time t+1

[@samlopezruiz]

Amazing series. Very clear explanations!

[@bengonoobiang6633]

Very interesting to understand the signal alignment. Thanks

[@dermaniac5205]

05:45 is this the right interpretation of alpha? Alpha is P(O1...Ot, qt=Si), which is the probability of observing O1..Ot AND being in state Si at timepoint t. But you said it is the probability of being in state Si at timepoint t GIVEN the Observations O1..Ot. That would P(qt=Si | O1...Ot) which is different.

[@iAmEhead]

Echoing what others have said... great videos, very useful. If you feel inclined I'd love to see some on other CS topics.

[@sahilgupta2210]

Well this was one of the best playlists I have gone through to pass my acads :) lol

[@edoardogallo9298]

WHAT A SERIES! that is a teacher..

[@djp3]

thanks!

[@arezou_pakseresht]

Thanks for the AMAZING playlist!

[@djp3]

Glad you like it!

[@sanketshah7670]

thank you so much for this....this is better than my ivy league tuition

[@djp3]

Glad it helped!

[@preetgandhi1233]

Very clear explanation, Mr. Ryan Reynolds....XD

[@myzafran1]

Thank you so much for your very clear explanation.

[@hariomhudiya8263]

That's some quality content, great series

[@djp3]

Glad you enjoy it!

[@benjaminbenjamin8834]

I wish Professor could also implement those concepts in python notebook also.

[@djp3]

there is a package called hmmlearn in conda-forge that has an implementation.

[@voxgun]

Thankyou so much for sharing Prof !

[@djp3]

You’re welcome!

[@teemofan7056]

Oh welp there goes 10000 of my brain cells.

[@djp3]

Hopefully 10,001 will grow in their place!

[@mindthomas]

Thanks for a thorough and well-taught video series. Is it possible to download the slides anywhere?

[@harikapatel3343]

You explained it so well.... thank you so much

[@Hugomove]

Great explained, thank you very very much!

[@djp3]

Glad it was helpful!

[@lakshmipathibalaji873]

Thanks for such a great explanation

[@djp3]

Glad it was helpful!

[@AmerAlsabbagh]

Your lectures are great, thanks, one note is that, beta is wrongly expressed in your video, and it should be the following: β is the probability of seeing the observations Ot+1 to OT, given that we are in state Si at time t and given the model λ, in other words, what is the probability of getting a specific sequence from a specific model if we know the current state.

[@djp3]

That sounds right. did I misspeak?

[@konradpietras8030]

@@djp3 In 7:00 u said that beta captures the probability that we would be in a givent state knowing what's going to come in the future. So it's the other way round, you should condition on current state not future observations.

[@timobohnstedt5143]

Excellent content. If I got it right, you state that the EM-algorithm is called gradient ascent or decent. If I got it right, this is not the same. The algorithms result can be in the same local optima, but they are not the same.

[@djp3]

if you abstract the two algorithms enough they are the same. But most computer scientists would recognize them as different algorithms that both find local optima.

[@minhtaiquoc8478]

Thank you for the lectures. The sound at the beginning and the end is really annoying though

[@parhammostame7593]

Great series! Thank you!

[@edwardlee6055]

I get through the vedio series and feel rescued.

[@Chi_Pub666]

You are the goat of teaching bw algorithm🎉🎉🎉

[@xntumrfo9ivrnwf]

"... 2 dimensional transition matrix (in principle)..." --> could anyone help with an example where e.g. a 3D transition matrix is used? Thanks.

[@djp3]

Moving through a skyscraper. Going from x,y,z to a new x,y,z

[@oriion22]

Hi Donald, Thanks for putting this easy to understand HMM series. I wanted to know a little bit more on how to apply it in other fields. How can I connect with you to discuss this.

[@djp3]

Twitter? @djp3

[@danilojrdelacruz5074]

Thank you and well explained!

[@djp3]

Glad you enjoyed it!

[@alikikarafotia4788]

Amazing series.

[@hayoleeo4891]

Thank you so much! I found it so hard to understand baum welch!

[@djp3]

You're very welcome!

[@anqiwei5784]

Wow! This video is so great!!!

[@djp3]

Thank you so much!!

[@naveenrajulapati3816]

Great explanation sir...Thank You

[@djp3]

You're most welcome

[@VishnuDixit]

Amazing playlist Thanks

[@quonxinquonyi8570]

Simply brilliant

[@punitkoujalgi7701]

You helped a lot.. Thank you

[@markusweis295]

Thank you! Nice video. (You look a bit like Ryan Reynolds)

[@djp3]

You think so? Amazon's automatic celebrity recognizer thinks I look like Shane Smith (at least with my beard)

[@threeeyedghost]

I was thinking the same for the whole video.

[@anqiwei5784]

Haha I think it's more than just a bit

[@lejlahrustemovic541]

You're a life saver!!!

[@fgfanta]

Quite the tour de force, thank you!

[@djp3]

ha!

[@akemap4]

One thing I cannot understand. If gamma is the sum of zeta over all j, then how can gamma have the dimension of T. If zeta only goes from 1 to T?

[@alexmckinney5761]

I noticed this too, it is better to use the alternate formulation for gamma, which is \gamma_t(i) = \alpha_t(i) * \beta_t(i) / \sum_i (\alpha_t(i) * \beta_t(i)). This should give you the correct dimension

[@djp3]

there is a matrix of gamma's for each t and each i and a 3-D matrix Xi's for each t,i,j. Each gamma_t is the sum over as set of Xi's at that time. You could also notate gamma as being gamma(t,i) and Xi and Xi(t,i,j)

[@akemap4]

@@alexmckinney5761 yes. I did it. However I still am getting error in my code. My a matrix goes to 1 on one side and zero on the other side. I am still trying to figure out the problem, but without success till then.

[@abdallahmahmoud8642]

Thank you! You are truly awesome

[@djp3]

You too!!

[@pauledson397]

Ahem: "ξ" ("xi") is pronounced either "ksee " or "gzee". You were pronouncing "xi" as if it were Chinese. But... still a great video on HMM and Baum-Welch. Thank you!

[@djp3]

Yes you are correct. I'm awful with my Greek letters.

[@toopieare]

Thank you professor!

[@sanketshah7670]

it seems you're mixing up gamma and delta?

[@djp3]

Possibly, do you mean the slides are wrong or I am misspeaking? I'm really bad with my Greek letters.

[@sanketshah7670]

@@djp3 no just delta is viterbi, not gamma, i think you say gamma is viterbi.

[@AakarshNair]

Really helpful

[@snehal7711]

greatttttt lecture indeed!

[@glassfabrikat]

Nice! Thank you!

[@djp3]

No problem

[@HuyNguyen-sn6kh]

you're a legend!

[@m_amirulhadi]

are u Deadpool?

[@kuysvintv8902]

I thought it's ryan reynolds

[@jiezhang3689]

ξ is pronounced as "ksaai"

[@djp3]

Yes. I pretty much botched that.

[@fjumi3652]

the ending :D :D :D

[@ozlemelih]

Who's she?

[@djp3]

?

[@TheCaptainAtom]

great video. pronounced 'ksi'.

[@djp3]

Yes. I totally blew that.