(ML 14.4) Hidden Markov models (HMMs) (part 1)

แชร์
ฝัง
  • เผยแพร่เมื่อ 7 ม.ค. 2025

ความคิดเห็น • 49

  • @shenaniganmusic57
    @shenaniganmusic57 8 ปีที่แล้ว +78

    Please don't whine about how he explains very slowly! It's his pace that he wishes to explain with. I listened to the tutorial at X1.25 speed and it is great.

  • @tetlleyplus
    @tetlleyplus 7 ปีที่แล้ว +3

    This guy is amazing,, the explanations are very clear, thanks!

  • @EDUARDO12348
    @EDUARDO12348 9 ปีที่แล้ว +2

    Thank you for proving a clear example. That helped a lot

  • @Turbo_TechnoLogic
    @Turbo_TechnoLogic 11 ปีที่แล้ว +2

    For everyone who does not understand the point of the HMM, just read the 'rainy-sunny' example on the Wikipedia.

  • @DDranks
    @DDranks 11 ปีที่แล้ว +1

    Big thanks for this video! I'm actually trying to understand Conditional Random Fields, a discrimitive (not generative) version of HMM... but this helped me get the basic terminology and facts straight! Thanks.

  • @jmrjmr8254
    @jmrjmr8254 10 ปีที่แล้ว +9

    Could you maybe show using the hand recognition software, what would the transitional probabilities, emission probabilities and initial distribution be?

  • @HassanMakkeh
    @HassanMakkeh 11 ปีที่แล้ว +4

    I would like to thank you very much for this great set of videos and your amazing way of explaining all those concepts. Loved your example about the handwritten recognition and the HMM. Thanks :)

  • @jaleelahammed
    @jaleelahammed 4 ปีที่แล้ว

    Nice Explanation, Really helpful, Thanks .

  • @mrshah2043
    @mrshah2043 7 ปีที่แล้ว +1

    Pretty interesting, thank you for this explanation.

  • @xinyixu8268
    @xinyixu8268 8 ปีที่แล้ว

    great video, very clear explanation!

  • @daniilsmirnov3153
    @daniilsmirnov3153 ปีที่แล้ว

    Interesting and easy to follow

  • @worldlover92
    @worldlover92 5 ปีที่แล้ว

    Thank you so much, for the video and explanation 👌👍

  • @lakeshed
    @lakeshed 12 ปีที่แล้ว

    itz called hi dden because the parameters that define the random variables are not seen. only the variables which are a different set are seen, example is: you are a researcher on weather, you want to find out what the summer looked like in 1852, you found a diary written in 1852, this person has written out how he spend money on that summer, you realized that he has spend pretty a lot of money buying ice creams and cold water for the family that summer, you would conclude that it was really hot

  • @emmanuelnsanga
    @emmanuelnsanga 11 ปีที่แล้ว +3

    Could you please do hierarchical hidden markov models?

  • @AlphaNovaImploder
    @AlphaNovaImploder 7 ปีที่แล้ว

    Loved your explanation and clear example, thanks!

  • @orochimarusama9499
    @orochimarusama9499 12 ปีที่แล้ว

    @mathematicalmonk: Nice video, but I have questions. This is the first time I am trying to understand what HMM's really mean. So when you say Z1...Zk are random variables which can assume any discrete values in {1,...,m}, I understand that Z symbolizes the hidden states but what is the physical (or graphical / intuitive) meaning of the set {1,..,m}. How are these related to the indices of Z (time steps/ state steps). An example might help. Thanks in advance

  • @naureenshaukat4875
    @naureenshaukat4875 8 ปีที่แล้ว

    nicely explained..... good job

  • @aronlindberg83
    @aronlindberg83 11 ปีที่แล้ว

    Question: what math is required to understand your treatment of HMMs?

  • @ablervoer8205
    @ablervoer8205 9 ปีที่แล้ว

    Thank you! What software did you used to record this drawing video with pen?

  • @jmrjmr8254
    @jmrjmr8254 10 ปีที่แล้ว +7

    I didn't quite understand the concept of emission probabilities. Anyone?

    • @mikel5264
      @mikel5264 6 ปีที่แล้ว

      It is the prob distribution of the observable for each hidden state.

    • @kallurisaikrishna862
      @kallurisaikrishna862 5 ปีที่แล้ว +3

      It is like this. In the given example lets say the first state Z_1 is 8. Then if we just take the letter correspodning to that number then X_1 will be H. However, we are implictly saying that if Z_1=8 then X_1=H with probability 1. This might not always be the case. There can be an error in the software sometime and in those cases even when the state Z_1=8, we can have X_1='K'. We can model this as given Z_1=H we have X_1='H' with probability 0.99 and X-1='K' with probability 0.01. These probabilities are called emission probabilities. Hence the dimension of emission probabilities is equal to the number of discrete observable variables possible (assuming the observations are discrete). I hope it is clear.

  • @nilspin
    @nilspin 10 ปีที่แล้ว +1

    Man this is so simple yet I don't quite understand 100% of HMM concept.
    Time to brush up prereq math, I guess. (:/)

  • @zabermi
    @zabermi 12 ปีที่แล้ว

    great work ! thanks for the video :)

  • @omendarine
    @omendarine 11 ปีที่แล้ว

    I guess that knowledge in probability (Bayes rule etc.) and some linear algebra (matrix multiplication) should be sufficient

  • @yuwang6841
    @yuwang6841 8 ปีที่แล้ว

    what is the soft in the radio?please help

  • @fanwang9005
    @fanwang9005 11 ปีที่แล้ว

    Really helpful!

  • @Alexithymiander
    @Alexithymiander 10 ปีที่แล้ว

    pmf stands for probability mass function :)

  • @jinkilove8020
    @jinkilove8020 8 ปีที่แล้ว

    just started learning HMM bcoz my prof chose this as my summer project can u suggest me some books i am kinda slow so i need it with lot of examples

  • @lixiao64
    @lixiao64 13 ปีที่แล้ว

    pretty good! clear !

  • @lakeshed
    @lakeshed 12 ปีที่แล้ว

    so the fact that he ate lotz of ice cream is an evidence that you did not witness thus hidden

  • @AfeezAzizFatboy
    @AfeezAzizFatboy 12 ปีที่แล้ว

    Thanks for the video!

  • @Turbo_TechnoLogic
    @Turbo_TechnoLogic 11 ปีที่แล้ว

    My pleasure ;)

  • @RaphaelShu
    @RaphaelShu 12 ปีที่แล้ว

    Thank a lot for the video

  • @WillWolfrick
    @WillWolfrick 13 ปีที่แล้ว

    Tnx for the video!

  • @google_cloud
    @google_cloud 9 ปีที่แล้ว +7

    Cut to the chase. You explain things too slowly. I keep fast forwarding to get to the important parts.

    • @pmcgee003
      @pmcgee003 9 ปีที่แล้ว +21

      +Anon Maybe you can still get your money back.

    • @majidaldo
      @majidaldo 9 ปีที่แล้ว +4

      +Anon i agree with you but i prefer this pace to allow things to sink in

    • @vasishtapolisetty639
      @vasishtapolisetty639 5 ปีที่แล้ว

      1.25x is optimum

  • @sengaun543
    @sengaun543 วันที่ผ่านมา

    B is in.
    I'm f hi us°'

  • @jesuslovesu4ever803
    @jesuslovesu4ever803 12 ปีที่แล้ว

    Didn`t understand the shit

  • @zarinaoflaz5495
    @zarinaoflaz5495 9 ปีที่แล้ว

    too slow

  • @drake1636
    @drake1636 7 ปีที่แล้ว

    wow what a terrible explanation