Questions about type in defineing problems

[Da1sypetals]

I do not exactly know how to conclude the question, so I am showing an example:
For example, I want to use the Newton solver, and I have to implement Gradient and Hessian.
I wonder what types can I use to be type Gradient? Am I allowed to use sparse matrices? Is this documented anywhere?

impl Gradient for Rosenbrock {
    type Gradient = /* what types are valid? */;
    // ... omitted
}

I browsed the docs and getting-started website but did not find anything related.
Thanks a lot in advance❤️

[stefan-k]

You can use any type you want; however, this type is required to implement a number of traits which depend on the solver you are using. The easiest way is to use one of the types supported by the various math backends (docs, book), such as Vec<f64>, Vec<f32>, ndarray::Array1<f64>, ndarray::Array1<f32>, and the various vectors supported by nalgebra.

If you choose to use your own type, you will have to implement the relevant traits on that type. For the Newton solver you can find the trait bounds listed here. In particular you would have to implement argmin_math::ArgminScaledSub for your type and argmin_math::ArgminInv and argmin_math::ArgminDot for an associated matrix type H.