search

SciML / DiffEqFlux.jl

Pre-built implicit layer architectures with O(1) backprop, GPUs, and stiff+non-stiff DE solvers, demonstrating scientific machine learning (SciML) and physics-informed machine learning methods
https://docs.sciml.ai/DiffEqFlux/stable
MIT License
871 stars 156 forks source link

UODE with neural networks. Beginner questions #831

Closed ghost closed 1 year ago

ghost commented 1 year ago

Hello developers,

I am new to Julia, UODE and sciML.

So I have a few basic questions which I would like to clarity:

  1. Where can I ask question related to sciML? Is the Issues section of this github appropriate?
  2. Are there any video tutorial on how to set up and train UODE's with the sciML package?
  3. Is there any example using neuralODE's with UODE? In the Lotka-Volterra system in https://github.com/ChrisRackauckas/universal_differential_equations/blob/master/LotkaVolterra/scenario_1.jl its appears to be a feed-forward NN (Or am I wrong?)
  4. What does this means in Julia nn_dynamics!(du,u,p,t) = ude_dynamics!(du,u,p,t,p_). A function with parameters equal to another function with parameters.

Thank you for developing this package. I have a problem with 5 dimensions (dx_1,dx_2,dx_3,dx_4, dx_5) where I only know part of the physics but have a tone of data.
This package seems to be right tool to model the information the physical model is not capturing.

Best Regards, James

prbzrg commented 1 year ago

Hi, Happy to see a newcomer

  1. For general questions, Julia forum is the best way to get help. And it has a section for specific domains. https://discourse.julialang.org/c/domain/10
  2. There is an online book that can help you see and learn tools and APIs of SciML. https://book.sciml.ai/
  3. To know the details of UODE, see the citations of this package. https://github.com/SciML/DiffEqFlux.jl/blob/master/CITATION.bib
  4. It's a way to define a function, To know more about functions in Julia, see the official manual of it at https://docs.julialang.org/en/v1/manual/functions/
ChrisRackauckas commented 1 year ago

Feel free to ask more questions, but closing since it seems answered.