Unexpected results from Brent minimizer.

[alexgian]

Unexpected results from Brent minimizer.

[sritchie]

I used this opportunity to create a nice visual debugging notebook in Maria, thanks to @mhuebert: https://2.maria.cloud/gist/e2e9fa24033f5b50d6cc7af44f22243f

The difference with scmutils was the initial condition for Brent. Apache Math-Commons used the midpoint between the left and right ends of the interval, while the original algorithm, and scmutils, used a golden section cut. It doesn't matter usually, but in this case the difference pushed you to a different local minimum.

I'm going to fix / change this to match scmutils in #175 . The real thing you want to use is #114 , and that would be an awesome contribution @alexgian if you're willing to work on it.

[alexgian]

Does this mean that in different circumstances choosing the midpoint between left and right would have been the better choice? Or is a Golden Section cut always preferred? If the the former is the case, then it needs be an option, I think. Yeah, I'll look at #114, Sam.

[sritchie]

@alexgian I think this is an edge case for the algorithm and there's no general answer here. I did provide a new option for :initial-guess, but I adjusted the default to the golden cut in #175 .

I'll close this now since this is fixed in #175 , and that's under review now by @littleredcomputer .

Thanks @alexgian !