Applet Orchard

Coding Challenge #100.4: Neuroevolution Flappy Bird – Part 4

[ad_1]
Part 4! I attempt to improve the Neuroevolution Flappy Bird Coding Challenge.

Part 1:
Part 2:
Part 3:
Part 4:
Part 5:

💻

🔗 Improved code:
🔗 TensorFlow.js:
🔗 ml5.js:
🔗 Flappy Bird Clone:
🔗 Neuroevolution documentation:

🎥 Coding Challenge #31: Flappy Bird:
🎥 My Neural Networks playlist:
🎥 My Genetic Algorithm playlist:

🚂 The Coding Train website:
💖 Support this channel on Patreon:
To buy Coding Train merchandise:
📚 Book recommendations:

🎥 For an Introduction to Programming:
🎥 For More Coding Challenges:

🔗
🔗

Posted

in

by

appletorchard

Tags:

Comments

20 responses to “Coding Challenge #100.4: Neuroevolution Flappy Bird – Part 4”

  1. KuraIthys Avatar
    KuraIthys

    Does kind of show how simple a game flappy bird is that it's possible to 'evolve' a solution with some of these versions in 8 generations or less. XD

    Reply
  2. Lucas Avatar
    Lucas

    WOW I actually got the idea from another comment, but It would really be amazing if you made a Rubik's cube first and its solver with a neural network if that is possible with your library !!!

    Reply
  3. TwinkyWinky Avatar
    TwinkyWinky

    Can you code a Rubik's Cube solver?

    Reply
  4. JinTsen Avatar
    JinTsen

    In the original you wouldn't die if you hit the floor. I used to go along the floor until I had to rise

    Reply
  5. X-Mux 01 Avatar
    X-Mux 01

    Nice Work!

    Reply
  6. René Carannante Avatar
    René Carannante

    3:40 inputs[4] = 0.5 + this.velocity / (.5 * height);

    Reply
  7. Tori Knorth Avatar
    Tori Knorth

    Honestly I'm impressed that the birds could do as well as they did when they didn't have the y-velocity information. It did occur to me that if the birds had a more complicated set of hidden nodes they could theoretically evolve to be able to calculate the y-velocity from successive y-value inputs though. This series of videos is really interesting.

    Edit: given that the y-velocity is the type of input that could theoretically be anything between minus infinity and infinity, it seems to me that it could be rescaled to be between -1 and 1 by plugging into a non-linear function like tanh(y-velocity). That's something I'd like to look into when I try doing neuro-evolution.

    Reply
  8. ArimB1000 Avatar
    ArimB1000

    you are the best coding channel thank you for taking your time to make videos for us

    Reply
  9. kosas Avatar
    kosas

    Kaputt xd

    Reply
  10. kustomweb Avatar
    kustomweb

    I think Saving and Loading brains is super important.

    Reply
  11. harshit jain Avatar
    harshit jain

    How can I contribute on github?

    Reply
  12. M4RF Avatar
    M4RF

    is it to late to start learning to code with 30 ?

    Reply
  13. Vikram Cothur Avatar
    Vikram Cothur

    I think the last change you made was more like flappy bird. It looked cooler too. Shouldn't have changed it back 🙁

    Reply
  14. moises fernandez Avatar
    moises fernandez

    Language?

    Reply
  15. germanLc Avatar
    germanLc

    i liked the video because 1:01

    Reply
  16. ScratchedRecord Avatar
    ScratchedRecord

    hey. I need help setting up a localhost. any chance you guys help? thanks

    Reply
  17. hypno flux Avatar
    hypno flux

    can you make a neural evolution game where there exists no winning or losing or being better?

    Reply
  18. kalapita226 Avatar
    kalapita226

    Thanks for video 🙂

    Reply
  19. Renan Machado Avatar
    Renan Machado

    Hi

    Reply
  20. jed moutahir Avatar
    jed moutahir

    Enfin, je l'attendait

    Reply

Leave a Reply

Your email address will not be published. Required fields are marked *