The implementation should follow the existing implementations.
Document the actual formula
For each test function there should also be functions which calculate the derivative and Hessian
Test functions (and derivative and Hessian) with a variable number of parameters should come in two forms, one where the number of variables is unknown at compile-time (possibly requires allocation on the heap) and a const generics version, which does not allocate.
Tests
Test the optimum
Test that derivative at the optimum is 0
Test that the derivative (roughly) matches the derivative obtained with finitediff
Test that the Hessian (roughly) matches the Hessian obtained with finitediff
Add benchmarks to benches/testfunctions.rs
Add function to python interface in ./python/argmin-testfunctions-py and add an example to the readme.
This should be a fairly easy thing to do and is therefore a great starting point for someone who wants to start contributing to argmin.
There are myriads of test functions which could be supported in
argmin_testfunctions
. Some sources are:A few thoughts:
finitediff
finitediff
benches/testfunctions.rs
./python/argmin-testfunctions-py
and add an example to the readme.This should be a fairly easy thing to do and is therefore a great starting point for someone who wants to start contributing to argmin.