(ML 18.8) Correctness of the Metropolis algorithm

Поделиться
HTML-код
  • Опубликовано: 11 сен 2024
  • Conditions under which the Metropolis algorithm is guaranteed to converge to the correct result.

Комментарии • 17

  • @tikke8511
    @tikke8511 7 лет назад +2

    Thank you very much on videos 18.x. After months of struggling, I finally understand the key idea of what makes MCMC to work. Especially the proof that Markov chain (Xi) has stationary distribution, put all the messy pieces into the right place in my head.

    • @bakih8468
      @bakih8468 7 лет назад

      Definitely a great series! I struggled for one day and after seeing these videos I understood what MCMC is all about. I guess I was lucky to have found this series so fast. Nice work!

  • @andyrichardson2434
    @andyrichardson2434 4 года назад +2

    I am struggling to understand what is being said around 2:30 to 3:10 minute mark. First in the audio he says x_i+1 NEQ x, but in the video he writes x_i NEQ x, so which is correct, audio or video? And then he goes on to reason why we can't have chosen the reject option, because then they would be equal, but the reject option is x_i+1 = x_i, not x_i+1 = x?

  • @6sixstringsdown6
    @6sixstringsdown6 12 лет назад +1

    Thank you. All of these videos are fantastic. I'm glad I watched them.

  • @charlesyin289
    @charlesyin289 8 лет назад

    I am really lucky to find this series of video

  • @user-zk4xv1gp9h
    @user-zk4xv1gp9h 2 года назад

    Thanks a lot!!! very helpful!!!!

  • @RealMcDudu
    @RealMcDudu 3 года назад

    great video

  • @jingliu4261
    @jingliu4261 6 лет назад +1

    What's the difference between Q and T (proposal vs transition matrix)? I thought when we sample x from Q(x_i, x), we are walking on a markov chain with transition Q.

    • @AP-rs5wz
      @AP-rs5wz 5 лет назад

      We are, but T is with the added rejection part.

  • @ASHARAHMAD799
    @ASHARAHMAD799 12 лет назад +1

    Great videos .Thankyou :-)

  • @TimBate
    @TimBate 11 лет назад

    Given that there is always a possibility of rejecting a transition and staying in out state for at least one state (and presumably many), then can't we simply say that if Q is irreducible then T is aperiodic, without having to show that Q is also aperiodic?

  • @googoonight858
    @googoonight858 6 лет назад

    very good explanation, thank you

  • @marcpan6750
    @marcpan6750 7 лет назад

    Great videos for MCMC.

  • @homataha5626
    @homataha5626 2 года назад

    where are u these days?

  • @muratcan__22
    @muratcan__22 5 лет назад

    this was great!

  • @holzermarco1
    @holzermarco1 11 лет назад

    rather charming

  • @EfratAbramovitz
    @EfratAbramovitz 8 лет назад

    You are so smart