17 June 2024 01:53:21 PM

ppc_neural_pde_square_hole():
  ppc_neural_pde() models a partial differential equation
  in the form of a boundary value problem
  over a square region with a hole,
  solved using a neural network.
  weights before training:
 -0.222 
  0.054 
 -0.023 
  0.129 
 -0.135 
  0.013 
  0.452 
  0.416 
  0.136 
  0.217 
 -0.358 
  0.107 
 -0.484 
 -0.257 
 -0.363 
  0.304 
 -0.343 
 -0.099 
 -0.370 
 -0.391 
  Nelder-Mead: Converged after 1407 function evaluations
  Nelder-Mead: Neural network’s residual error = 3.21515e-14
  weights after training:
 -2.225 
  6.291 
  1.707 
 -1.406 
 -1.776 
  5.474 
  2.116 
  2.347 
  0.704 
 -4.816 
 -5.313 
 -0.608 
 -4.374 
 -1.170 
 -0.576 
  1.133 
  5.462 
 -0.995 
  0.944 
 -1.303 
  Error versus the ODE’s exact solution = 31.5

ppc_neural_ode_square_hole():
  Normal end of execution.
17 June 2024 01:53:21 PM