Hidden Markov Model : Data Science Concepts

[@13_yashbhanushali40]

Unbelievable Explanation!! I have referred to more than 10 videos where basic working flow of this model was explained but I must say that rather I'm sure that this is the most easiest explanation one can ever find on youtube , the way of explanation considering the practical approach was much needed and you did exactly that Thanks a ton man !

[@TesfaldetBokretsion]

True experts always make it easy.

[@pinkymotta4527]

Crystal-clear explanation. Didn't have to pause video or go back at any point of video. Would definitely recommend to my students.

[@chadwinters4285]

I have to say you have an underrated way of providing intuition and making difficult to understand concepts really easy.

[@mohammadmoslemuddin7274]

Glad I found your videos. Whenever I need some explanation for hard things in Machine Learning, I come to your channel. And you always explain things so simply. Great work man. Keep it up.

[@ritvikmath]

Glad to help!

[@paulbrown5839]

To get to the probabilities in the top right of the board, you keep applying P(A,B)=P(A|B).P(B) ... eg. A=C3, B=C2 x C1 x M3 x M2 x M1 ... keep applying P(A,B)=P(A|B).P(B) and you will end up with same probabilities as shown on the whiteboard top right of screen for the viewer. Great video!

[@ritvikmath]

Thanks for that!

[@ummerabab8297]

Sorry, but I still don't get the calculation at the end. The whole video was explained flawlessly but the calculation was left out. I don't understand. If you can please further help. Thankyou.

[@toyomicho]

@@ummerabab8297 Here is some code in python showing the calculations in the output, you'll see that the hidden sequence s->s->h has the highest probability (0.018) ##### code #################### def get_most_likely(): starting_probs={'h' :.4, 's':.6} transition_probs={'hh':.7, 'hs':.3, 'sh':.5, 'ss':.5, } emission_probs = {'hr':.8, 'hg':.1,'hb':.1, 'sr':.2, 'sg':.3, 'sb':.5} mood={1:'h', 0:'s'} # for generating all 8 possible choices using BitMasking observed_clothes = 'gbr' def calc_prob(hidden_states:str)->int: res = starting_probs[hidden_states[:1]] # Prob(m1) res *= transition_probs[hidden_states[:2]] # Prob(m2|m2) res *= transition_probs[hidden_states[1:3]] # Prob(m3|m2) res *= emission_probs[hidden_states[0]+observed_clothes[0]] # Prob(c1|m1) res *= emission_probs[hidden_states[1]+observed_clothes[1]] # Prob(c2|m2) res *= emission_probs[hidden_states[2]+observed_clothes[2]] # Prob(c2|m3) return res #Use BitMasking to generate all possible combinations of hidden states 's' and 'h' for i in range(8): hidden_states = [] binary = i for _ in range(3): hidden_states.append(mood[binary&1]) binary //=2 hidden_states = "".join(hidden_states) print(hidden_states, round(calc_prob(hidden_states),5)) ##### Output ###### sss 0.0045 hss 0.0006 shs 0.00054 hhs 0.000168 ssh 0.018 hsh 0.0024 shh 0.00504 hhh 0.001568

[@AakashOnKeys]

@@toyomicho I had the same doubt. Thanks for the code! Would be better if author pins this.

[@zishiwu7757]

Thank you for explaining how HMM model works. You are a grade saver and explained this more clearly than a professor.

[@ritvikmath]

Glad it was helpful!

[@stevengreidinger8295]

You gave the clearest explanation of this important topic I've ever seen! Thank you!

[@remy4033]

This guy is underrated for real. Love you bro.

[@coupmd]

Wonderful explanation. I hand calculated a couple of sequences and then coded up a brute force solution for this small problem. This helped a lot! Really appreciate the video!

[@rssamarth099]

This helped me at the best time possible!! I didn't know jack about the math a while ago, but now I have a general grasp of the concept and was able to chart down my own problem as you were explaining the example. Thank you so much!!

[@clauzone03]

You are great! Subscribed with notification after only the first 5 minutes listening to you! :-)

[@ritvikmath]

Aw thank you !!

[@marceloamado6223]

You are a great professor! Thank you very much for taking the time to make this video all the best to you.

[@beyerch]

Really great explanation of this in an easy to understand format. Slightly criminal to not at least walk through the math on the problem, though.

[@nathanielfernandes8916]

I have 2 questions: 1. The Markov assumption seems VERY strong. How can we guarantee the current state only depends on the previous state? (e.g., person has an outfit for the day of the week instead of based on yesterday) 2. How do we collect the transition/emission probabilities if the state is hidden?

[@straft5759]

1. It is strong, but the idea is that each state (at least in principle) encodes *all* the information you need, i.e. the entire "memory" of the system. So for example, if the person's mood tomorrow depends on their mood yesterday as well as today, then you would model that as a 4-state system (HH, HS, SH, SS) instead of a 2-state system (H, S). 2. This problem in particular assumes that you already know those probabilities, but if you didn't you could still Bayesian them out of the collected data. That's more advanced though.

[@awalehmohamed6958]

Instant subscription, you deserve millions of followers

[@songweimai6411]

Really appreciate your work. Much better than the professor in my class who has a pppppphhhhdddd degree.

[@Dima-rj7bv]

I really enjoyed this explanation. Very nice, very straightforward, and consistent. It helped me to understand the concept very fast.

[@ritvikmath]

Glad it was helpful!

[@jirasakburanathawornsom1911]

Im continually amazed by how well and easy to understand you can teach, you are indeed an amazing teacher

[@mirasan2007]

Dear ritvik, I watch your videos and I like the way you explain. Regarding this HMM, the stationary vector π is [0.625, 0.375] for the states [happy, sad] respectively. You can check the correct stationary vector by multiplying it with the transpose of the Transition probability Matrix, then it should result the same stationary vector as result: import numpy as np B = np.array([[0.7, 0.3], [0.5, 0.5]]) pi_B = np.array([0.625, 0.375]) np.matmul(B.T, pi_B) array([0.625, 0.375])

[@caspahlidiema4027]

The best ever explanation on HMM

[@ritvikmath]

thanks!

[@ahokai]

I don't know why I had paid for my course and then came here to learn. Great explanation, thank you!

[@mengxiaoh9048]

thanks for the video! I've watched two other videos but this one is the easiest to understand HMM and I also like that you added the real-life application NLP example at the end

[@ritvikmath]

Glad it was helpful!

[@totomo1976]

Thank you so much for your clear explanation!!! Look forward to learning more machine-learning related math.

[@1243576891]

This explanation is concise and clear. Thanks a lot!

[@ritvikmath]

Of course!

[@ananya___1625]

As usual awesome explanation...After referring to tons of videos, I understood it clearly only after this video...Thank you for your efforts and time

[@ritvikmath]

You are most welcome

[@Molaga]

A great video. I am glad I discovered your channel today.

[@ritvikmath]

Welcome aboard!

[@Infaviored]

If there is a concept I did not understand from my lectures, an i see there is a video by this channel, i know I will understand it afterwards.

[@ritvikmath]

thanks!

[@Infaviored]

@@ritvikmath no, thank you! Ever thought of teaching at an university?

[@pibob7880]

After watching this it left me with the impression that local maximization of conditional probabilities lead to global maximization of the hidden markov model. Seems too good to be true... I guess the hard part is finding out the hidden state transition probabilities?

[@VascoDaGamaOtRupcha]

You explain very well!

[@ashortstorey-hy9ns]

You're really good at explaining these topics. Thanks for sharing!

[@louisc2016]

I really like the way you explain something, and it helps me a lot! Thx bro!!!!

[@shivkrishnajaiswal8394]

Nice explanation!! One of the usecases mentioned was NLP. I am wondering if HMM will be helpful given that we now have Transformers architectures.

[@slanglabadang]

I feel like this is a great model to use to understand how time exists inside our minds

[@gopinsk]

I agree Teaching is an art. You have mastered it. Application to real world scenarios are really helpful. Really feel so confident after watching your videos. Question, How did we get the probabilities to start with? are those arbitrary or followed any scientific method to arrive at those numbers?

[@OskarBienko]

I'm curious too. Did you figure it out?

[@mia23]

Thank you. That was a very impressive and clear explanation!

[@ritvikmath]

Glad it was helpful!

[@seansanyal1895]

hey Ritvik, nice quarantine haircut! thanks for the video, great explanation as always. stay safe

[@ritvikmath]

thank you! please stay safe also

[@linguipster1744]

oooh I get it now! Thank you so much :-) you have an excellent way of explaining things and I didn’t feel like there was 1 word too much (or too little)!

[@spp626]

Such a great explanation! Thank you sir.

[@Aoi_Hikari]

i had to rewind the videos a few times, but eventually i understood it, thanks

[@froh_do4431]

really good work on the simple explanation of a rather complicated topic 👌🏼💪🏼 thank you very much

[@kristiapamungkas697]

You are a great teacher!

[@ritvikmath]

Thank you! 😃

[@kanchankrishna3686]

Why are there 8 possible combinations (6:10)? I got 9 from doing M1/G, M1/B, M1/R, M2/G, M2/B, M2/R, M3/G, M3/R, M3/B ?

[@laurelpegnose7911]

Great video to get an intuition for HMMs. Two minor notes: 1. There might be an ambiguity of the state sad (S) and the start symbol (S), which might have been resolved by renaming one or the other 2. About the example configuration of hidden states which maximizes P: I think this should be written as a tuple (s, s, h) rather than a set {s, s, h} since the order is relevant? Keep up the good work! :-)

[@jinbowang8814]

Really nice explanation! easy and understandable.

[@skyt-csgo376]

You're such a great teacher!

[@shahabansari5201]

Very good explanation of HMM!

[@ritvikmath]

Glad it was helpful!

[@hichamsabah31]

Very insightful. Keep up the good work.

[@arungorur3305]

Ritvik, great videos.. I have learnt a lot.. thx. A quick Q re: HMM. How does one create transition matrix for hidden states when in fact you don't know the states.. thx!

[@kiran10110]

Damn - what a perfect explanation! Thanks so much! 🙌

[@ritvikmath]

Of course!

[@kanhabansal524]

best explanation over internet

[@ritvikmath]

Thanks!

[@claytonwohl7092]

At 2:13, the lecturer says, "it's not random" whether the professor wears a red/green/blue shirt. Not true. It is random. It's random but dependent on the happy/sad state of the professor. Sorry to nitpick. I definitely enjoyed this video :)

[@ritvikmath]

Fair point !! Thanks :)

[@ls09405]

Great Video. But how did you calculate {SSH} is maximum?

[@5602KK]

Incredible. All of the other videos I have watched have me feeling quite over whelmed.

[@ritvikmath]

glad to help!

[@srijanshovit844]

Awesome explanation I understood in 1 go!!

[@newwaylw]

Why are we maximizing the joint probability? Shouldn't the task to find the most likely hidden sequence GIVEN the observed sequence? i.e. maximizing the conditional probability argmax P(m1m2m3| c1c2c3)?

[@otixavi8882]

Great video, however I was wondering if the hidden state transitioning probabilities are unknown, is there a way to compute/calculate them based on the observations?

[@mihirbhatia9658]

I wish you went through Bayes Nets before coming to HMM. That would make the conditional probabilities so much more easier to understand for HMMs. Great explanation though !! :)

[@sarangkulkarni8847]

Absolutely Amazing

[@mansikumari4954]

This is great!!!!!

[@beckyb8929]

beautiful! Thank you for making this understandable

[@shaoxiongsun4682]

Thanks a lot for sharing. It is very clearly explained. Just wondering why the objective we want to optimize is not the conditional probability P(M=m | C = c).

[@Justin-General]

Thank you, please keep making content Mr. Ritvik.

[@NickVinckier]

This was great. Thank you!

[@ritvikmath]

Glad you enjoyed it!

[@souravdey1227]

Really crisp explanation. I just have a query. When you say that the mood on a given day "only" depends on the mood the previous day, this statement seems to come with a caveat. Because if it "only" depended on the previous day's mood, then the Markov chain will be trivial. I think what you mean is that the dependence is a conditional probability on the previous day's mood: meaning, given today's mood, there is a "this percent" chance that tomorrow's mood will be this and a "that percent" chance that tomorrow's mood will be that. "this percent" and "that percent" summing up to 1, obviously. The word "only" somehow conveyed a probability of one. I hope I am able to clearly explain.

[@GarageGotting]

Fantastic explanation. Thanks a lot

[@ritvikmath]

Most welcome!

[@alecvan7143]

Very insightful, thank you!

[@anna-mm4nk]

appreciate that the professor was a 'she' took me by surprise and made me smile :) also great explanation, made me remember that learning is actually fun when you understand what the fuck is going on

[@user-or7ji5hv8y]

This is really great explanation

[@paulbrown5839]

@ritvikmath Any chance of a follow up video covering some of the algos like Baum-Welch, Viterbi, please? ... i'm sure you could explain them well. Thanks a lot.

[@ritvikmath]

Good suggestion! I'll look into it for my next round of videos. Usually I'll throw a general topic out there and use the comments to inform future videos. Thanks!

[@juanjopiconcossio3146]

Great great explanation. Thank you!!

[@srinivasuluyerra7849]

Great video, nicely explained

[@gnkk6002]

Wonderful explanation 👌

[@ritvikmath]

Thank you 🙂

[@silverstar6905]

verry nice explanation. looking forward to seeing something about quantile regression

[@Sasha-ub7pz]

Thanks, amazing explanation. I was looking for such video but unfortunately, those authors have bad audio.

[@froh_do4431]

Is it possible to describe in a few words, how we can calculate/compute the transition- and emission probabilities?

[@wendyqi4727]

I love your videos so much! Could you please make one video about POMDP?

[@PeteThomason]

Thank you, that was a very clear introduction. They key thing I don't get is where the transition and emission probabilities come from. In a real-world problem, how do you get at those?

[@jordanblatter1595]

In the case of the NLP example with part of speech tagging, the model would need data consisting of sentences that are assigned tags by humans. The problem is that there isn't much of that data lying around.

[@SuperMtheory]

Great video. Perhaps a follow up will be the actual calculation of {S, S, H}

[@ritvikmath]

thanks for the suggestion!

[@dariocline]

I'd be flipping burgers without ritvikmath

[@mousatat7392]

amazing keep up very cool explenation

[@ritvikmath]

Thanks!

[@minapagliaro7607]

Great explanation ❤️

[@ResilientFighter]

Ritvik, it might be helpful if you add some practice problems in the description

[@froh_do4431]

What is the most common algorithm used, to maximize the probabilities? ...just to give a hint on this part of the whole model

[@user-or7ji5hv8y]

How did you factorize the joint into conditionals? Is there a link?

[@MegaJohnwesly]

oh man. Thanks alot :). I tried to understand here and there by reading..But I didn't get it. But this video is gold

[@ritvikmath]

Glad it helped!

[@chia-chiyu7288]

Very helpful!! Thanks!

[@ritvikmath]

Glad it was helpful!

[@user-or7ji5hv8y]

Can you matrix multiply transmission with emission since they look like matrices?

[@tindo0038]

here is my quick implementation of the discussed problem index_dict = {"happy": 0, "sad": 1} start_prob = {"happy": 0.4, "sad": 0.6} transition = [[0.7, 0.3], [0.5, 0.5]] emission = { "happy": {"red": 0.8, "green": 0.1, "blue": 0.1}, "sad": {"red": 0.2, "green": 0.3, "blue": 0.5}, } observed = ["green", "blue", "red"] cur_sequece = [] res = {} def dfs(cur_day, cur_score): if cur_day >= len(observed): res["".join(cur_sequece)] = cur_score return cur_observation = observed[cur_day] for mood in ["happy", "sad"]: new_score = cur_score new_score += emission[mood][cur_observation] # at the start, there is no previous mood if cur_sequece: new_score += transition[index_dict[mood]][index_dict[cur_sequece[-1]]] else: new_score += start_prob[mood] cur_sequece.append(mood) dfs(cur_day + 1, new_score) cur_sequece.pop() dfs(0, 0) print(res)

[@zacharyzheng3610]

Brilliant explanation

[@ritvikmath]

Thanks!

[@curiousredpand90]

Ah you explained so much better than my Ivy League professor!!!

[@AnkeetKumar]

Where does one get these probabilites from?

[@VIJAYALAKSHMIJ-h2b]

good explanation. But the last part of determining the moods is left out. How did you get s,s,h

[@RezaShokrzad]

BIG LIKE, Absolutely awesome. just could you explain about the interpretation of {SSH}? Should we compute all 8 cases of m_i, then compare them?

[@ritvikmath]

Thanks! And yes exactly, we can do that. In practice, of course with many time periods and states this gets too expensive so we have more efficient ways to compare them but at the end of the day we are still getting the maximum.

[@hmyswonderland4532]

great video! but i was wondering why the p(C2|m3,m2,m1)..., why the m3 is related to the c2?

[@qiushiyann]

Thank you for this explanation!

[@SPeeDKiLL45]

Great Video Bro ! Thanks

[@lallawmsangachhangte2949]

Can you post a video on POS tagging with CRF please

[@user-or7ji5hv8y]

Cool. Have you done a video on how to get those probabilities from observed data? Is it using MCMC?

[@ingoverhulst]

Great work! I really enjoy your content.

[@TylerHNothing]

for the lexical example of determining adjective verb etc, I'm guessing some training dataset had to be compiled to list each dictionary word and the frequency of part of speech. just doesn't seem feasible that I would have training data about how often a particular word is adjective vs verb etc but then need to build a classifier. there's not many problems I can think of where this is useful... how could I have training data where the class is not hidden, but then have the class be hidden in the future? Someone had to ask the teacher about happy/sad etc to get the Emission probabilities, why can't they ask her again?