This is an absolutely great channel. i was reading page rank in the university and thought i understood markhov chains well, but upon watching your video i realised there was so much more intuitive things to learn, Thanks a lot!!!
Thank you for your Video, I just want to know some more details of the last part. In how many iterations of the Markov chain do you expect that see the specific random walk? As we know by increasing the iteration of random walks it would be more chance to appear spesific random walk. can you please clear this part?
Method 2 curls my nails a bit. Instead of of multiplying A many times you could just square A again and again, which will lead to A^(2^n) after n steps. That would be so much faster.
Just a question. If I would like to find outlier sequences, say, the ones with a very low probability, it is clear that I just need to use your last example and calculate some kind of percentile threshold on my dataset of probabilities of chains to isolate the outliers. I am dealing with sequences of different length, so, bigger sequences will naturally tend to 0 due to the fact that probabilities are between 0 and 1. The question is... I am using geometric mean of the transitions, this is, after multiplying all the probabilities like you do, I calculate the nth root of that product, this to get a "better value" that is not affected by the chain length (in this case n where n can be between "a" and b" for some a,b) in order to be able to compare them, is this the correct approach for this problem ? Thanks
I'm trying to be helpful here, but it is difficult to make out a lot of your ode for a couple of reasons. First, even on a relatively large monitor, the font is so small I have a hard time making it out. Solution: You have a lot of negative space to the right of your code. Try bringing the screen in closer. Second: the dark mode screen obfuscates certain colors of type. It would be helpful if you used a different color mode or increased the size of the code relative to the screen. Other than that, great job!
This is an absolutely great channel. i was reading page rank in the university and thought i understood markhov chains well, but upon watching your video i realised there was so much more intuitive things to learn, Thanks a lot!!!
Excellent! The eigenvector perspective is cool - shows how they, once again, reveal the most important characteristics of a system.
Yeah...exactly!
Lol just solved this problem for my code right before you posted this video! Your explanation would have saved me a decent amount of time
Haha...it happens XD
Absolutely great, useful and easy to follow. Many thanks!
Man your outro is awesome!
Great video! Greetings from Argentina
Thanks mate! :D
Great Video! Thank you
Supose that we a huge A matrix (realy big dimension). Would be faster/cheaper to use the nontecarlo aproach rather than finding the eigenvectors?.
Thank you for your Video, I just want to know some more details of the last part. In how many iterations of the Markov chain do you expect that see the specific random walk? As we know by increasing the iteration of random walks it would be more chance to appear spesific random walk. can you please clear this part?
can you make video on HMM practical implementation please
How do you determine the probability of the transition matrix?
To quote South Park:
"Pizza, french fries, pizza, french fries....see? he should of done pizza instead of french fries."
Great video!
I have a doubt. I think the probability function should be divided by "len (seq)", do I?
How could I plot a graph of the random walk over each iteration?
Method 2 curls my nails a bit. Instead of of multiplying A many times you could just square A again and again, which will lead to A^(2^n) after n steps. That would be so much faster.
Good point, but at this level of matrix dimensionality it computes fast enough anyway...
Haha i was going to comment that.
Just a question. If I would like to find outlier sequences, say, the ones with a very low probability, it is clear that I just need to use your last example and calculate some kind of percentile threshold on my dataset of probabilities of chains to isolate the outliers. I am dealing with sequences of different length, so, bigger sequences will naturally tend to 0 due to the fact that probabilities are between 0 and 1. The question is... I am using geometric mean of the transitions, this is, after multiplying all the probabilities like you do, I calculate the nth root of that product, this to get a "better value" that is not affected by the chain length (in this case n where n can be between "a" and b" for some a,b) in order to be able to compare them, is this the correct approach for this problem ? Thanks
is it possible to trade NIFTY india using MARKOV chain ? mostly on the LONG side?
U r the best
Keep supporting :D
I don't know if you mentioned, but google's search algorithm uses these stationary distributions!!
Yeahh!...No, I didn't mention that in this video.
can you make a full course on stats and coding? i would be wiling to pay for the service. thanks for the content.
Is there a way of verifying the values of stationary state?
He's just shown you three methods. If they agree, no problem. In an earlier vid if the system is oscilatory the stationary state will not settle.
can you make practical implementation video on HMM please
I can has Baum-Welch algo next pls?
I'm trying to be helpful here, but it is difficult to make out a lot of your ode for a couple of reasons. First, even on a relatively large monitor, the font is so small I have a hard time making it out. Solution: You have a lot of negative space to the right of your code. Try bringing the screen in closer. Second: the dark mode screen obfuscates certain colors of type. It would be helpful if you used a different color mode or increased the size of the code relative to the screen. Other than that, great job!
Thanks for the feedback!
Last time it was a hamburger and not a hot dog😂
Really? 😂😂
Hi. How to do it for time series data?
Very good point, especially where the states have long lives or variation time between or in transition
How to run this
Hi. Awesome videos! Which college are you/were you from?
Kalyani Government Engineering College.
Why didn't you use FOR loops instead of WHILE loops?
Haha...no specific reason