rishi-s8 / algorithmicdifferentiation-spark Goto Github PK

Add the necessary details about the implementations. Add more details on how to use the module.

Problem
The Tangentify decorator is needed for the functions inside the Tangent_Type class. Moreover, the decorator type-casts into the Tangent_Type Class. Therefore the Tangentify decorator is written twice.

Solution
No idea.

Minor corrections needed.

Problem:
The classes Tangent_Mode and Tangent_Mode_Distributed in Math.py have common code in their functions, including the constructor.

Solution:
Use class inheritance

Problem
There are no tests for the module, hence it is very hard to check the correctness of any changes introduced.

Solution
Add Pytest tests.

The existing implementation can be extended to calculate the Jacobian using the adjoint model computation.

Problem
In each function, the incoming variable is first typecasted to the required class.
Hence, there is the same piece of code at the start of every function.

Solution:
Use Python decorators.

Currently, the tangent is being calculated in Math.py. Create 2 separate modules for Tangent_Mode and Tangent_Mode_Distributed and put in a src.

We need to have a better name for the repository. :)

I was thinking something like:

• SparkyAD
• BigADPy

Need more suggestions.

Should we make the repository public? It would be easier to include CI apps to the repository if we have the repo public. @Manvi07