Bro dropped the coldest playlist on Markov Chains and thought we wouldn't notice 🥶 . Sir 🙏🙏🙇♂🙇♂. These animation combined videos just make you understand 30 mins topic in 10 mins.
I've only just discovered this channel and watched the Markov Chains series. They're excellent. They explained the concept and practice clearly and relatively simply. I'll definitely be coming back to see what else is available here. It's a great resource.
Thank you so much for the informative video. I am enrolled in a machine learning course with no background in calculus or linear algebra and this video is literally a life saver
Your presentation and teaching quality gives a positive vibe specially your voice. Moreover great to see an Indian teaching in such a beautiful way. Highly recommended channel who wish to learn advance probability theory. Have a great smoothness in your teaching. Thank you Normalizing Nerd
Thank you so much for the videos. You explained Markov Chain a lot better than my professor. You have mentioned aperiodicity in this video but you have not talked about it in later videos.
This is the best HMM series in youtube. Crystal clear explanation and the music in between the video is just superb... If you can send me the link to the music would be great. Thanks in advance.
Notes for my future review. 05:06 *Chanpman-Kolmogorov Theorem* can be used to calculate the probability of going from state i to state j after n steps. 05:06 = Probability of going from state i to state j in n steps with an intermiadte stop in state k after r steps, then sum over all possible value of k. Probability for going from state i to state j after n step = Probililty of going from state i into a in-between state * Probability of going from the in-between states to j state, for all the possible in-between state = ( Proability of going from state i to k after r steps * Probaility of going from k to n after n-r steps ).sum_for_all_k_states In other words: Get the probability of going from state-i to state-j in n step, through a state k. Do it for all possible state k. Then sum all the probabilities. But what is r? What is the value of r? The system reaches an intermidiate state k after r steps. There are different possible k states, so for each k, the r (the number of step to reach from i to k) might be different. The exact value of r is not important, except that probability of k to j is for n-r steps.
Amazing Content :) For any who wondered like my why getting from 0 to 2 through 1 in 2 steps is the product of getting from 0 to 1 with 1 to 2, it follows from the Chapman Kolomogorov Thm explained around 5:54 :)
Thank you. I am doing a course in probability models and I have never seen Markov Chains. The textbook examples aren't the most intuitive so it has been a bit frustrating. This helps a lot.
Hello, future person here. I really like your videos regarding markov chains :) I was wondering if it be possible to also get the mathematical definitions that go along with the videos. Will you also be covering markov decision programs and go into detail regarding optimal value function/policy? It would also be really interesting to have a parallel series on continuous state (and continuous action if we are talking about decision programs) markov chains as i dont know any good sources who cover them!
Humm interesting...Markov decision process and it's use in RL is a broad topic. Currently, I don't possess a very good understanding of them. Maybe someday I will.
Very explicative and informative, concise and to the point. Please add more videos about the applications of Chapman-Kolmogorov in variant applied fields.
When you brought up Chapman-Kolmogorov theorem proof and said “you can go through it” this felt slightly off putting, because I couldn't. I mean, everything else is so clearly stated, and looks like I should've understood the theorem proof too, but failed to. I dunno whether I'd feel better if the proof was explained or if I knew what exactly the proof notation means or if it was in the description. Anyway, your channel is great, you're so good at breaking complicated stuff down to simple stuff, I just wanted to give some feedback
Thanks for your honest feedback. Well, I didn't explain the derivation because that kinda bores people. Here's a tip to understand the derivation: take any two states i and j, n=4. Now try to compute P_ij(n) using r=0,1,2,3. You'll see how things work.
@ Itachi Uchiha Assuming num of states = 3 Suppose you have taken r=1 so, P11(3) = P11(1) * P11(2) + P12(1) * P21(2) + P13(1) * P31(2) Again applying CK theorem on 2 step transitions, P11(2) = P11(1) * P11(1) + P12(1) * P21(1) + P13(1) * P31(1) P21(2) = P21(1) * P11(1) + P22(1) * P21(1) + P23(1) * P31(1) P31(2) = P31(1) * P11(1) + P32(1) * P21(1) + P33(1) * P31(1) Similarly, you can use any value of r
Nice channel, I am waiting for this excellent explanation. Next, how about addressing the Attention Mechanism. Thank you so much for your excellent channel.
Just wanted to say there's been a little tiny error at ~4:30. It's stated that P_{ij}(n) = A_{ij}^n, that's not exactly correct in index notation, that's the equivalent of raising each individual element to a power and not the actual matrix. For n=2, P_{ij}(2) = A_{ik}A_{kj}, n=3 P_{ij}(3) = A_{ik}A_{kl}A_{lj}. It's a little more convoluted, but it would simply work the way you want with addition of parenthesis. P_{ij}(n) = (A^n)_{ij}
Great videos! Could you please clarify whether we just use the pi values we found in the first video to find A^infinity as I recall that in the first video we found that pi=[0.3591 0.2145 0.43564] which are not the same as 0.444 0.333 or 0.222 (which are the values of A^infinity). Thank you!
Because of your explanations on matrices, during my exam instead of solving all the pis for the transition matrix, i just made my calculator power it to 100 to answer it correctly on my exam 😂😂😂
The content is good, but you probably should put your mouth closer to the microphone because I turn up the volume to hear your voice but then the ad in the video blows up my headphones.
