todo.md

TODO list

Last work : DeepINN/constraint/gradients.py

Geometry

• Geometry added but lot of bloats from TorchPhysics. Clean up geometry folder. In progress in "DeepINN/geometry_refactor".
• Clean up utils folder.
• Implement anchor points.
• Add prediction plot function.

Gradients

• Implement basic gradients.
• Implement gradient for multiple output neurons.
• Do we need retain_graph=True?

Constraints

• Implement the prescribed BC part in constraint/ boundary_loss dirichletBC.
• Implement PDE loss constraint.
• Implement gradients.
• Implement lazy evaluation of gradients.
• Implement more constraint.

Architectures

• Implement fully connected NN.
• Implement more neural networks

Tutorials

• Basic geometry tutorials.
• Constraints tutorials.
• Add template PDE in constraint directory.
• There was some problem with FCNN tutorial. It was incorrect PDE.
• Add 2D discontinuous heat conduction problem
• Add Fourier neural network and Deep Galerkin method architecture.

Misc

• Migrate to JupyterBooks.
• Move everything after contribution in the readme.md to the docs.

Technical stuff

• Substitute raise NotImplementedError to @abc.abstractmethod for better ABC implementation on concrete classes (classes where the member function are actually implemented).
• Use functool.wraps so that the decorator doesn't loose the metadata such as docsting.