Descent methods and line search: preconditioned steepest descent

Поделиться
HTML-код
  • Опубликовано: 18 ноя 2024

Комментарии • 17

  • @__amkhrjee__
    @__amkhrjee__ День назад

    Extremely well made video! It'd be nice if you make a tutorial on how to make those nice plots.

  • @avk8477
    @avk8477 7 месяцев назад

    Extremely concise and lucid explanation. Thank you Prof. Michel.

  • @tanweermahdihasan4119
    @tanweermahdihasan4119 4 года назад +5

    Lovely explanation! Please continue teaching, your style of coaching is commendable.

  • @sephgeodynamics9246
    @sephgeodynamics9246 3 года назад +2

    Thank you :)

  • @Makstuff
    @Makstuff 2 года назад

    What I'm confused about: At 1:41 you say we follow the direction _until the function starts increasing again_, but looking at the table at 3:35 I see a constant value alpha=0.2. Which I'd interpret considering the bottom formula at 8:17 as that you don't follow the direction *until the function starts increasing again*, but rather follow it a fixed ratio of 0.2 times the gradient length?

    • @MichelBierlaire
      @MichelBierlaire  2 года назад +1

      In this example, it just happens that the step to perform to reach the minimum is always the same.

  • @zhengyangli8474
    @zhengyangli8474 3 года назад

    Really nice explanation! For the preconditioning, is there any method to determine matrix H?

    • @MichelBierlaire
      @MichelBierlaire  3 года назад

      You can find various review papers in the literature.
      doi.org/10.1017/S0962492915000021
      doi.org/10.1006/jcph.2002.7176

    • @zhengyangli8474
      @zhengyangli8474 3 года назад

      @@MichelBierlaire Thank you, Prof. Michel.

  • @castellsmatrix2078
    @castellsmatrix2078 3 года назад +2

    which software do you use?

  • @miqomargaryan15
    @miqomargaryan15 6 месяцев назад

    kroasan