Available on crate feature regression only.
Expand description

A random binary searching n-variable optimizer.

Independently binary searches the variables over the entire range of f64s. Supports any number of variables to be optimized together.

This has performance in the ballpark of OLS, but enables you to give your own function, which means you can optimize for things other than least squares along straight lines. This opens up the opportunity to fit other functions (any you want) and to use functions less prone to outliers (least squares is very prone).


  • Options for the binary search optimization.
  • Dynamically sized storage, for use when the number of variables isn’t known at compile time.


  • A trait which allows storage of n-variable optimization, either on the stack through arrays ([f64; VARIABLE_COUNT]) or allocated on the heap through Vec.