Struct GradientDescent 

pub struct GradientDescent<S = NoSearch, Q = BarzilaiBorwein> {
    pub abs_tol: Tolerances,
    pub dual: bool,
    pub error_norm: Norm,
    pub line_search: S,
    pub max_steps: usize,
    pub rel_tol: Option<Scalar>,
    pub step_estimator: Q,
}

The method of gradient descent.

Fields

abs_tol: Tolerances

Absolute error tolerances.

dual: bool

Lagrangian dual.

error_norm: Norm

Norm type for error evaluation.

line_search: S

Line search algorithm.

max_steps: usize

Maximum number of steps.

rel_tol: Option<Scalar>

Relative error tolerance.

step_estimator: Q

Step size estimator.

Implementations

impl GradientDescent

pub fn default() -> Self

impl<S, Q> GradientDescent<S, Q>

pub fn with_line_search<T>(self, line_search: T) -> GradientDescent<T, Q>

The same solver, searching along each step with another line search.

pub fn with_step_estimator<T>(self, step_estimator: T) -> GradientDescent<S, T>

The same solver, estimating each step size with another estimator.

Trait Implementations

impl<S, Q> Debug for GradientDescent<S, Q>

where S: ObjectiveMerit, Q: StepEstimator,

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl<S> Default for GradientDescent<S>

where S: Default,

fn default() -> Self

Returns the “default value” for a type.

impl<S, Q, F, J, X, E> Optimization<F, J, (), X> for GradientDescent<S, Q>

where S: ObjectiveMerit, Q: StepEstimator, F: Erase<Erased = Scalar> + Tensor, <J as Tensor>::Unit: UnitMul<<X as Tensor>::Unit>, <<J as Tensor>::Unit as UnitMul<<X as Tensor>::Unit>>::Output: UnitSum, <<<J as Tensor>::Unit as UnitMul<<X as Tensor>::Unit>>::Output as UnitSum>::Output: Is<<F as Tensor>::Unit>, J: Erase<Erased = E> + Jacobian + Mul<StepSize<J, X>, Output = X>, for<'a> &'a J: Mul<StepSize<J, X>, Output = X>, X: Erase<Erased = E> + Jacobian + Solution, <X as Tensor>::Unit: UnitDiv<<J as Tensor>::Unit>, E: Tensor, for<'a> &'a Matrix: Mul<&'a X, Output = Vector>,

fn minimize( &self, function: impl FnMut(&X) -> Result<F, String>, jacobian: impl FnMut(&X) -> Result<J, String>, _hessian: impl FnMut(&X) -> Result<(), String>, initial_guess: X, equality_constraint: EqualityConstraints, _sparse: Option<SparseSolver>, ) -> Result<X, OptimizationError>

impl<Q, F, X, E> RootFinding<F, (), X> for GradientDescent<NoSearch, Q>

where Q: StepEstimator, F: Erase<Erased = E> + Jacobian + Mul<StepSize<F, X>, Output = X>, for<'a> &'a F: Mul<StepSize<F, X>, Output = X>, X: Erase<Erased = E> + Jacobian + Solution, <X as Tensor>::Unit: UnitDiv<<F as Tensor>::Unit>, E: Tensor, for<'a> &'a Matrix: Mul<&'a X, Output = Vector>,

fn root( &self, function: impl FnMut(&X) -> Result<F, String>, _jacobian: impl FnMut(&X) -> Result<(), String>, initial_guess: X, equality_constraint: EqualityConstraints, _sparse: Option<SparseSolver>, ) -> Result<X, OptimizationError>