Struct DlrBuilder 

pub struct DlrBuilder<S>
where
    S: StatisticsType,
{ /* private fields */ }

Builder for the independent (interpolative-decomposition) DLR.

use sparse_ir::{DlrBuilder, Fermionic};
let dlr = DlrBuilder::<Fermionic>::new(10.0, 10.0)
    .accuracy(1e-10)
    .build()
    .unwrap();
assert!(!dlr.poles().is_empty());

Implementations

impl<S> DlrBuilder<S>

where S: StatisticsType + 'static,

pub const DEFAULT_ACCURACY: f64 = 1e-14

Default target accuracy, 1e-14.

The poles are selected by a pivoted Gram–Schmidt factorization of the kernel in double precision, whose residuals cannot fall much below 1e-15. An accuracy below about 1e-14 is therefore beyond what the selection can resolve: it keeps adding near-redundant poles chosen by rounding, which makes the representation larger and its fits less accurate. 1e-14 is the tolerance the reference implementations use at full precision (cppdlr warns below it).

pub fn new(beta: f64, wmax: f64) -> DlrBuilder<S>

Start a DLR for inverse temperature beta and frequency cutoff wmax.

pub fn accuracy(self, accuracy: f64) -> DlrBuilder<S>

Target relative accuracy of the kernel interpolation.

Values below about 1e-14 are beyond double-precision resolution; see Self::DEFAULT_ACCURACY.

pub fn max_size(self, max_size: usize) -> DlrBuilder<S>

Upper bound on the number of poles.

pub fn build(self) -> Result<DiscreteLehmannRepresentation<S>, Error>

Select the poles and build the representation.

Errors

Error::InvalidParameter if beta or wmax is not positive and finite, the accuracy is not in (0, 1), or max_size is 0.

Trait Implementations

impl<S> Clone for DlrBuilder<S>

where S: Clone + StatisticsType,

fn clone(&self) -> DlrBuilder<S>

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl<S> Debug for DlrBuilder<S>

where S: Debug + StatisticsType,

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

Formats the value using the given formatter.