Question about latent_ode.py noise_std?

rtqichen / torchdiffeq

Differentiable ODE solvers with full GPU support and O(1)-memory backpropagation.

MIT License

5.61k stars 930 forks source link

Question about latent_ode.py noise_std? #212

Open Yunyi-learner opened 2 years ago

Yunyi-learner commented 2 years ago

On line 266, when computing the loss, why can we use noise_std directly, I mean noise_std is a pre-defined parameter for us to generate our original (true) trajectories, when we calculate loss, we should not use any pre-defined true parameter, is that correct?