Solving Newton's Law of Cooling with Physics Informed Neural Networks (PINNs)

Поділитися
Вставка
  • Опубліковано 28 гру 2024

КОМЕНТАРІ • 14

  • @elastropy
    @elastropy  3 місяці тому +1

    Win the source code used in this video - www.elastropy.com/more/unlock-free-source-codes
    Join our Telegram group for exclusive access to detailed discussions, resources, programming files used in the video, and extra support! It's all free-click the link below to join now. See you there!
    Telegram Group Link - telegram.me/elastropy_official

  • @AdilDarvesh-w5f
    @AdilDarvesh-w5f 2 місяці тому

    Good understanding for me❤😊

    • @elastropy
      @elastropy  2 місяці тому

      Hi @AdilDarvesh-w5f, Thank you for your support! 😊 If you're interested in more content like this, feel free to check out my other tutorials in this playlist: ua-cam.com/video/7Es1ZWiMq0Y/v-deo.html&pp=gAQBiAQB.
      I hope you find them helpful!

  • @SumitKumar-qi2vc
    @SumitKumar-qi2vc 3 місяці тому

    Can this method be similar for non linear ode's

  • @EmmanuelOseiTutu-n7v
    @EmmanuelOseiTutu-n7v 3 місяці тому +1

    Excellent
    Much appreciated for your commitment

    • @elastropy
      @elastropy  3 місяці тому

      Hi @EmmanuelOseiTutu-n7v, thank you so much for your kind words!

  • @patelpavan5479
    @patelpavan5479 3 місяці тому +1

    Why multiply 4 and 2 in loss can we multiply less values like 0.1,0.2 , what is beneficial less or more weight value?

    • @elastropy
      @elastropy  3 місяці тому

      Hi @patelpavan5479, yes, you can assign weights lower than 1 in PINNs. The weights, like 4 and 2 in my video, are just random numbers used to balance the loss terms (reasons explained in the video). The weights control the balance between different loss terms, so smaller weights reduce the importance of a term. Just ensure the weights don’t downplay key components too much. It's all about balancing based on your specific problem, so feel free to experiment!
      Let me know if you have any more questions, and feel free to join our Telegram group for more updates and discussions!

    • @patelpavan5479
      @patelpavan5479 3 місяці тому

      @@elastropy thanks and make video on pde also.

    • @patelpavan5479
      @patelpavan5479 3 місяці тому

      With neumman type boundry conditions

    • @elastropy
      @elastropy  3 місяці тому

      Hi @patelpavan5479 Thanks for your suggestion! I'm planning to cover more topics on PDEs soon. Neumann boundary conditions will definitely be included! Stay tuned for upcoming videos.

    • @patelpavan5479
      @patelpavan5479 3 місяці тому

      @@elastropy Sure!!

  • @ramsaran_india
    @ramsaran_india 3 місяці тому +1

    Sir how can i satisfy exact initial condition like for my problem I have initial condition T(0) =0 ,T(1)= 1 but when I am predicting values at these conditions, I am not getting exact value so how can i modify my model.

    • @elastropy
      @elastropy  3 місяці тому

      Hi @ramsaran_india, Thank you for your question! I assume your domain is from 0 to 1, and you're trying to satisfy the initial and boundary conditions T(0)=0 and T(1)=1. If you're not getting the exact values at these points, one approach to address this is by adjusting the loss function, as I demonstrated in the tutorial.
      Specifically, you can give more weight to the loss terms that account for the initial and boundary conditions. This ensures that the model focuses more on matching these conditions during training. I encourage you to revisit the part of the video where I explain the multi-weighted loss function, as it can help with situations like this. Let me know if you have any more questions!