twiddle algorithm for parameter optimization (very simple haha)

The twiddle algorithm is used for parameter optimization of an algorithm. This corresponds to finding (hopefully) the global minimum of the loss function.


However, it does not work with gradients etc., it's just implemented by updating the parameters into one direction and getting feedback from the algorithm (did the error increase or decrease).


Great and short article with python code: https://martin-thoma.com/twiddle/

My implementation of Twiddle: https://github.com/moritzmitterdorfer/twiddle-parameter-optimizing/blob/master/README.md