David Schäfer
David Schäfer
  • 1
  • 8 254
NEAT Algorithm Visually Explained
NeuroEvolution of Augmenting Topologies (NEAT) is a genetic algorithm (GA) for training artificial neural networks based on concepts taken from evolutionary biology. This animation was made in python using the package 'manim'. The voiceover was created using ElevenLabs.
Music by Vincent Rubinetti.
Переглядів: 8 292

Відео

КОМЕНТАРІ

  • @Eeealt
    @Eeealt 13 днів тому

    Is (bestbirdw1 + 2ndbestbirdw1) / 2 = new w1?

  • @harshithdr3952
    @harshithdr3952 Місяць тому

    Wonderful explanation, keep making videos it’s really helpful for beginners

  • @dewaeq
    @dewaeq 2 місяці тому

    the production quality is amazing, wow!

  • @optozorax
    @optozorax 3 місяці тому

    Now try dense multilayer NN, and evolve it using CMA-ES algorithm, and you will be surprised how fast it will converge.

  • @medooazmi
    @medooazmi 3 місяці тому

    Thats neat

  • @zix2421
    @zix2421 5 місяців тому

    I love this algorithm

  • @hazardousharmonies
    @hazardousharmonies 5 місяців тому

    This is a superb demo David. Are you able to share your github code? Thanks for the video!

  • @revimfadli4666
    @revimfadli4666 5 місяців тому

    sounds quite unusual of NEAT to take this long for a simple problem Great video though!

  • @kotreq5862
    @kotreq5862 6 місяців тому

    Great explaination! Looking forward to more videos

  • @KX36
    @KX36 6 місяців тому

    "more of a lucky coincidence than a well thought out strategy" Or as it's called when Space-X boosters use this exact strategy intentionally in real life, a "suicide burn".

  • @fareahrahman1908
    @fareahrahman1908 6 місяців тому

    Please make more videos on different algorithms.

  • @mahdirafati6014
    @mahdirafati6014 7 місяців тому

    It was Awsome! Tahnk you.

  • @ExodusVFX
    @ExodusVFX 7 місяців тому

    Bless you in Jesus name

  • @RonicTheEgg
    @RonicTheEgg 8 місяців тому

    why u not become the famous how

  • @xDLiker
    @xDLiker 8 місяців тому

    Do I understand that correct, that one always needs to define a reward function (in this case score) for this type of problem?

    • @noahsmith673
      @noahsmith673 8 місяців тому

      yes

    • @medooazmi
      @medooazmi 3 місяці тому

      (Reward function = score function = fitness function) Basicly all the same, and they are just to tell the algorithm if it was a bad neural network, or a good little neurie :)

  • @nemongames9871
    @nemongames9871 8 місяців тому

    It might be out of context, but you have the the voice of OmniMan

  • @MrOnlineCoder
    @MrOnlineCoder 9 місяців тому

    Really great visuals, sad it has so small amount of views

  • @floppaplatinum5386
    @floppaplatinum5386 11 місяців тому

    This is really great for beginners! Keep it up. I would love to see other games and configurations for the algorithm and how they could affect the simulation e.g. faster training time etc.

  • @__henrik5474
    @__henrik5474 11 місяців тому

    Amazing explanations!