Enum SmoothFunctional 

pub enum SmoothFunctional<'a> {
    PointEvaluation {
        design_row: ArrayBase<ViewRepr<&'a f64>, Dim<[usize; 1]>>,
    },
    AverageDerivative {
        derivative_design: ArrayBase<ViewRepr<&'a f64>, Dim<[usize; 2]>>,
        weights: Option<ArrayBase<ViewRepr<&'a f64>, Dim<[usize; 1]>>>,
    },
    Contrast {
        design_row_a: ArrayBase<ViewRepr<&'a f64>, Dim<[usize; 1]>>,
        design_row_b: ArrayBase<ViewRepr<&'a f64>, Dim<[usize; 1]>>,
    },
    AverageValue {
        value_design: ArrayBase<ViewRepr<&'a f64>, Dim<[usize; 2]>>,
        weights: Option<ArrayBase<ViewRepr<&'a f64>, Dim<[usize; 1]>>>,
    },
    Linear {
        gradient: ArrayBase<ViewRepr<&'a f64>, Dim<[usize; 1]>>,
    },
}

Functional descriptor for the Layer-1 closed-form path. All variants are linear in the fitted coefficient vector.

Variants

PointEvaluation

m(x0), represented by the prediction/design row at x0.

Fields

design_row: ArrayBase<ViewRepr<&'a f64>, Dim<[usize; 1]>>

AverageDerivative

mean_i w_i * d m(x_i) / d x_j, represented by derivative-basis rows.

Fields

derivative_design: ArrayBase<ViewRepr<&'a f64>, Dim<[usize; 2]>>

weights: Option<ArrayBase<ViewRepr<&'a f64>, Dim<[usize; 1]>>>

Contrast

m(x_a) - m(x_b), represented by the two prediction design rows.

Fields

design_row_a: ArrayBase<ViewRepr<&'a f64>, Dim<[usize; 1]>>

design_row_b: ArrayBase<ViewRepr<&'a f64>, Dim<[usize; 1]>>

AverageValue

mean_i w_i * m(x_i), represented by value-basis rows.

Fields

value_design: ArrayBase<ViewRepr<&'a f64>, Dim<[usize; 2]>>

weights: Option<ArrayBase<ViewRepr<&'a f64>, Dim<[usize; 1]>>>

Linear

Direct caller-supplied linear functional gradient.

Fields

gradient: ArrayBase<ViewRepr<&'a f64>, Dim<[usize; 1]>>

Implementations

impl<'a> SmoothFunctional<'a>

pub fn gradient( &self, ) -> Result<ArrayBase<OwnedRepr<f64>, Dim<[usize; 1]>>, EstimationError>