Trait Amplitude

pub trait Amplitude:
    DynClone
    + Send
    + Sync
    + Serialize
    + Deserialize {
    // Required methods
    fn register(
        &mut self,
        resources: &mut Resources,
    ) -> Result<AmplitudeID, LadduError>;
    fn compute(
        &self,
        parameters: &Parameters<'_>,
        event: &Event,
        cache: &Cache,
    ) -> Complex<Float>;

    // Provided methods
    fn precompute(&self, event: &Event, cache: &mut Cache) { ... }
    fn precompute_all(&self, dataset: &Dataset, resources: &mut Resources) { ... }
    fn compute_gradient(
        &self,
        parameters: &Parameters<'_>,
        event: &Event,
        cache: &Cache,
        gradient: &mut DVector<Complex<Float>>,
    ) { ... }
    fn central_difference_with_indices(
        &self,
        indices: &[usize],
        parameters: &Parameters<'_>,
        event: &Event,
        cache: &Cache,
        gradient: &mut DVector<Complex<Float>>,
    ) { ... }
}

This is the only required trait for writing new amplitude-like structures for this crate. Users need only implement the register method to register parameters, cached values, and the amplitude itself with an input Resources struct and the compute method to actually carry out the calculation. Amplitude-implementors are required to implement Clone and can optionally implement a precompute method to calculate and cache values which do not depend on free parameters.

Required Methods

fn register( &mut self, resources: &mut Resources, ) -> Result<AmplitudeID, LadduError>

This method should be used to tell the Resources manager about all of the free parameters and cached values used by this Amplitude. It should end by returning an AmplitudeID, which can be obtained from the Resources::register_amplitude method.

fn compute( &self, parameters: &Parameters<'_>, event: &Event, cache: &Cache, ) -> Complex<Float>

This method constitutes the main machinery of an Amplitude, returning the actual calculated value for a particular Event and set of Parameters. See those structs, as well as Cache, for documentation on their available methods. For the most part, Events can be interacted with via Variables, while Parameters and the Cache are more like key-value storage accessed by ParameterIDs and several different types of cache IDs.

Provided Methods

fn precompute(&self, event: &Event, cache: &mut Cache)

This method can be used to do some critical calculations ahead of time and store them in a Cache. These values can only depend on the data in an Event, not on any free parameters in the fit. This method is opt-in since it is not required to make a functioning Amplitude.

fn precompute_all(&self, dataset: &Dataset, resources: &mut Resources)

Evaluates Amplitude::precompute over ever Event in a Dataset.

fn compute_gradient( &self, parameters: &Parameters<'_>, event: &Event, cache: &Cache, gradient: &mut DVector<Complex<Float>>, )

This method yields the gradient of a particular Amplitude at a point specified by a particular Event and set of Parameters. See those structs, as well as Cache, for documentation on their available methods. For the most part, Events can be interacted with via Variables, while Parameters and the Cache are more like key-value storage accessed by ParameterIDs and several different types of cache IDs. If the analytic version of the gradient is known, this method can be overwritten to improve performance for some derivative-using methods of minimization. The default implementation calculates a central finite difference across all parameters, regardless of whether or not they are used in the Amplitude.

In the future, it may be possible to automatically implement this with the indices of registered free parameters, but until then, the Amplitude::central_difference_with_indices method can be used to conveniently only calculate central differences for the parameters which are used by the Amplitude.

fn central_difference_with_indices( &self, indices: &[usize], parameters: &Parameters<'_>, event: &Event, cache: &Cache, gradient: &mut DVector<Complex<Float>>, )

A helper function to implement a central difference only on indices which correspond to free parameters in the Amplitude. For example, if an Amplitude contains free parameters registered to indices 1, 3, and 5 of the its internal parameters array, then running this with those indices will compute a central finite difference derivative for those coordinates only, since the rest can be safely assumed to be zero.

Trait Implementations

impl<'typetag> Serialize for dyn Amplitude + 'typetag

fn serialize<S>(&self, serializer: S) -> Result<S::Ok, S::Error>

where S: Serializer,

Serialize this value into the given Serde serializer.

impl<'typetag> Serialize for dyn Amplitude + Send + 'typetag

fn serialize<S>(&self, serializer: S) -> Result<S::Ok, S::Error>

where S: Serializer,

Serialize this value into the given Serde serializer.

impl<'typetag> Serialize for dyn Amplitude + Send + Sync + 'typetag

fn serialize<S>(&self, serializer: S) -> Result<S::Ok, S::Error>

where S: Serializer,

Serialize this value into the given Serde serializer.

impl<'typetag> Serialize for dyn Amplitude + Sync + 'typetag

fn serialize<S>(&self, serializer: S) -> Result<S::Ok, S::Error>

where S: Serializer,

Serialize this value into the given Serde serializer.

Implementors