Open chenyv118 opened 1 year ago
Please see new code at https://github.com/lu-group/sbinn
Please see new code at https://github.com/lu-group/sbinn
Yes that's right, my code didn't have a restriction on the division operation of state variables, using this new code solved the problem, thanks a lot!
I am trying to solve the ODE equation based on a model that modifies the glycolysis reaction, but the loss function keeps reporting nan and I can't get the result, even if I set the loss_weights of the ODE to 0, it doesn't help, the first 15 of the 17 state variables have a loss of nan. I hope someone can help me find the cause of the problem if possible.
The ODE equation of the system is as follows.
The code for data generation is as follows: Biodiesel_ODE.py
The neural network code is as follows: Biodisel_NN.py
I can sort out the parameters in k_config and data_config later if needed. One thing to note is that Fa will change after 120 minutes and will equal 0 when Fa>Vreactf, and this mutation may have some impact. I'm not sure.