At the moment the codebase uses a custom gradient descent algorithm - it is derived from the paper written by Baker and Matthews.
There are however many good options available in the scipy.optimize sub-package. We should try using these different methods and make them available to the user, for example as discussed here:
At the moment the codebase uses a custom gradient descent algorithm - it is derived from the paper written by Baker and Matthews.
There are however many good options available in the scipy.optimize sub-package. We should try using these different methods and make them available to the user, for example as discussed here:
https://github.com/nfaggian/python-register/issues/36