Enum ceres_solver::loss::LossFunction

pub enum LossFunction {
    Custom(CustomLossFunction),
    Stock(StockLossFunction),
}

Loss function for ResidualBlock and CurveFitProblem1D, it is a transformation of the squared residuals which is generally used to make the solver less sensitive to outliers. This enum has two flavours: user specified function and Ceres stock function.

Variants

Custom(CustomLossFunction)

User-specified loss function.

Stock(StockLossFunction)

One of the loss functions specified by Ceres, see Ceres Solver docs for details.

Implementations

impl LossFunction

pub fn custom(func: impl Into<LossFunctionType>) -> Self

Create a LossFunction to handle a custom loss function.

Arguments

• func - a boxed function which accepts two arguments: non-negative squared residual, and an array of 0) loss function value, 1) its first, and 2) its second derivatives. See details at http://ceres-solver.org/nnls_modeling.html#_CPPv4N5ceres12LossFunctionE.

pub fn huber(a: f64) -> Self

Huber loss function, see details at http://ceres-solver.org/nnls_modeling.html#_CPPv4N5ceres9HuberLossE.

pub fn soft_l1(a: f64) -> Self

Soft L1 loss function, see details at http://ceres-solver.org/nnls_modeling.html#_CPPv4N5ceres12SoftLOneLossE.

pub fn cauchy(a: f64) -> Self

log(1+s) loss function, see details at http://ceres-solver.org/nnls_modeling.html#_CPPv4N5ceres10CauchyLossE.

pub fn arctan(a: f64) -> Self

Arctangent loss function, see details at http://ceres-solver.org/nnls_modeling.html#_CPPv4N5ceres10ArctanLossE.

pub fn tolerant_loss(a: f64, b: f64) -> Self

Tolerant loss function, see details at http://ceres-solver.org/nnls_modeling.html#_CPPv4N5ceres12TolerantLossE.

pub fn loss(&self, squared_norm: f64, out: &mut [f64; 3])

Calls the underlying loss function.