A library for solving differential equations using neural networks based on PyTorch, used by multiple research groups around the world, including at Harvard IACS.
Currently, the loss function is assumed to be a function of the reparameterized output of the neural net alone. What if the user want a loss function that also takes the independent variable directly as an input?
Currently, the loss function is assumed to be a function of the reparameterized output of the neural net alone. What if the user want a loss function that also takes the independent variable directly as an input?