ConvolutionalNeuralOperators

This package implements Convolutional Neural Operators following this. The CNOs can then be used as custom Lux models and they are compatible with closure modeling.

Install

using Pkg
Pkg.add(url="git@github.com:DEEPDIP-project/ConvolutionalNeuralOperator.jl.git")

Usage

You probably want to use the cno function to create a closure model, which can be used in CoupledNODE or as a Lux model.

  closure, θ_start, st = cno(
      T = T,
      N = N,
      D = D,
      cutoff = cutoff,
      ch_sizes = ch_,
      activations = act,
      down_factors = df,
      k_radii = k_rad,
      bottleneck_depths = bd,
      rng = rng,
      use_cuda = false,
  )

to get the closure model, and then use it as a Lux model, or in CoupledNODE

  l, trainstate = CoupledNODE.train(
      closure,
      θ,
      st,
      dataloader,
      loss;
      tstate = trainstate,
      nepochs = 2,
      alg = Adam(T(1.0e-3),
  )

Look in test/ for more detailed examples on how to use the package, or look at the documentation.

How to Cite

If you use ConvolutionalNeuralOperators.jl in your work, please cite using the reference given in CITATION.cff.

Contributing

If you want to make contributions of any kind, please first that a look into our contributing guide directly on GitHub or the contributing page on the website

Contributors