Struct smoothed_online_convex_optimization::cost::CostFn

pub struct CostFn<'a, T, C, D>(_);

Cost functions that arrived over time. Individual cost functions may have different domains. For example, in a predictive online setting, a cost function arriving at time $t$ generally has the domain $[t, t + w]$.

Implementations

impl<'a, T, C, D> CostFn<'a, T, C, D> where
    T: Clone,
    C: ModelOutputSuccess,
    D: ModelOutputFailure, 

pub fn empty() -> Self

Creates empty cost function.

pub fn new(t: i32, f: SingleCostFn<'a, T, C, D>) -> Self

Creates initial cost function from some time $t \geq 1$.

pub fn add(&mut self, t: i32, f: SingleCostFn<'a, T, C, D>)

Adds a new cost function which may return uncertain predictions. Must always return at least one sample (which corresponds to certainty).

pub fn call_mean(&self, t: i32, x: T) -> Cost<C, D>

Returns mean if cost function returns a prediction.

pub fn call_certain(&self, t: i32, x: T) -> Cost<C, D>

Computes certain cost.

pub fn call_predictive(&self, t: i32, x: T) -> Vec<Cost<C, D>>

Computes uncertain cost.

pub fn call_mean_within_bounds<B>(&self, t: i32, x: T, bounds: &B) -> Cost<C, D> where
    B: DecisionSpace<'a, T>, 

Returns mean if cost function returns a prediction while ensuring that the given parameter is within the decision space.

pub fn call_certain_within_bounds<B>(
    &self,
    t: i32,
    x: T,
    bounds: &B
) -> Cost<C, D> where
    B: DecisionSpace<'a, T>, 

Computes certain cost while ensuring that the given parameter is within the decision space.

pub fn call_predictive_within_bounds<B>(
    &self,
    t: i32,
    x: T,
    bounds: &B
) -> Vec<Cost<C, D>> where
    B: DecisionSpace<'a, T>, 

Computes uncertain cost while ensuring that the given parameter is within the decision space.

Trait Implementations

impl<'a, T: Clone, C: Clone, D: Clone> Clone for CostFn<'a, T, C, D>

fn clone(&self) -> CostFn<'a, T, C, D>

Returns a copy of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl<'a, C, D> DiscretizableCostFn<'a, C, D> for CostFn<'a, FractionalConfig, C, D> where
    C: ModelOutputSuccess + 'a,
    D: ModelOutputFailure + 'a, 

fn into_i(self) -> CostFn<'a, IntegralConfig, C, D>

Discretize a certain fractional cost function.

impl<'a, C, D> RelaxableCostFn<'a, C, D> for CostFn<'a, IntegralConfig, C, D> where
    C: ModelOutputSuccess + 'a,
    D: ModelOutputFailure + 'a, 

fn into_f(self) -> CostFn<'a, FractionalConfig, C, D>

Relax a certain integral cost function to the fractional setting.

impl<'a, T, C, D> ResettableCostFn<'a, T, C, D> for CostFn<'a, T, C, D> where
    T: Clone,
    C: ModelOutputSuccess,
    D: ModelOutputFailure, 

fn reset(&'a self, t_start: i32) -> CostFn<'a, T, C, D>

Shift a certain cost function to some new initial time $t_start$ (a time before first time slot).