I used the NeuralPDE to solve the ODE system by following the tutorial at NeuralPDE.jl. I used the construction with the Lux.jl package, which is presented at Defining Systems of PDEs for Physics-Informed Neural Networks (PINNs).
In the Direct Construction via symbolic_discretize section, the loss function can take apart the pieces and reassemble them using the symbolic_discretize interface, and there is a difference between the construction of Lux and Flux packages. I would like to understand the network construction of two of them, but the tutorial just provides the following information: " Notice here that the solution is represented in the OptimizationSolution with u as the parameters for the trained neural network. But, for the case where the neural network is from Lux.jl, it’s given as a ComponentArray where res.u.depvar.x corresponds to the result for the neural network corresponding to the dependent variable x , i.e. res.u.depvar.u1 are the trained parameters for phi[1] in our example. For simpler indexing, you can use res.u.depvar[:u1] or res.u.depvar[Symbol(:u,1)] as shown here.". Would you please provide me the references for these two packages when using the symbolic_discretize interface?
Thank you all.