Struct ceres_solver::cost::CostFunction

pub struct CostFunction<'a>(_);

A cost function for ResidualBlock of the NllsProblem.

Implementations

impl<'a> CostFunction<'a>

pub fn new(
    func: impl Into<CostFunctionType<'a>>,
    parameter_sizes: impl Into<Vec<usize>>,
    num_residuals: usize
) -> Self

Create a new cost function for ResidualBlock from a Rust function.

Arguments

• func - function to find residuals and Jacobian for the problem block. The function itself must return false if it cannot compute Jacobian, true otherwise, and accept following arguments:
  • parameters - slice of f64 slices representing the current values of the parameters. Each parameter is represented as a slice, the slice sizes are specified by parameter_sizes.
  • residuals - mutable slice of f64 for residuals outputs, the size is specified by num_residuals.
  • jacobians: JacobianType - represents a mutable structure to output the Jacobian. Sometimes the solver doesn’t need the Jacobian or some of its components, in this case the corresponding value is None. For the required components it has a 3-D shape: top index is for the parameter index, middle index is for the residual index, and the most inner dimension is for the given parameter component index. So the size of top-level Some is defined by parameter_sizes.len(), second-level Some’s slice length is num_residuals, and the bottom-level slice has length of parameter_sizes[i], where i is the top-level index.
• parameter_sizes - sizes of the parameter vectors.
• num_residuals - length of the residual vector, usually corresponds to the number of data points.

pub fn parameter_sizes(&self) -> &[usize]

Lengths of the parameter vectors.

pub fn num_residuals(&self) -> usize

Length of the residual vector.

pub fn num_parameters(&self) -> usize

Parameter count.

pub fn cost(
    &self,
    parameters: &[&[f64]],
    residuals: &mut [f64],
    jacobians: JacobianType<'_>
) -> bool

Calls underlying cost function.