Struct nblast::ScoreMatrixBuilder

pub struct ScoreMatrixBuilder<T: TargetNeuron> { /* private fields */ }

Calculate a score matrix (lookup table for converting point matches into NBLAST scores) from real data using some sets of matching and non-matching neurons.

At a minimum, one matching set must be added (.add_matching_set), and the dist and dot bins must either be set automatically or calculated (.set_(n_)?(dist|dot)_bins).

Implementations

impl<T: TargetNeuron + Sync> ScoreMatrixBuilder<T>

pub fn new(neurons: Vec<T>, seed: u64) -> Self

seed is used when randomly selecting non-matching neurons to use as controls.

pub fn add_matching_set(&mut self, matching: &[usize]) -> &mut Self

Add a set of neurons which are considered to be mutually matching, as indices into the ScoreMatrixBuilder.neurons. This can be done several times with different sets. Matching sets should be small subsets of the total neuron count.

Indices outside of the neuron vec will be silently filtered out. Filtered sets with length <2 will also be silently ignored.

pub fn add_nonmatching_set(&mut self, nonmatching: &[usize]) -> &mut Self

Optionally list neuron sets which are mutually non-matching, as indices into ScoreMatrixBuilder.neurons, from which non-matching pairs will be randomly drawn. By default, all neurons are included, as matching sets are expected to be only a small portion of the total neurons.

Candidate non-matching pairs are made sure not to be in the matching pairs, so this can safely be ignored.

pub fn set_use_alpha(&mut self, use_alpha: bool) -> &mut Self

Alpha values are a measure of how colinear a neuron’s local neighborhood is. Score matrices calculated using alpha can only be used for queries using alpha, and vice versa.

False by default.

pub fn set_dist_lookup(&mut self, lookup: BinLookup<Precision>) -> &mut Self

Manually set the lookup bins for distances.

Consider BinLookup::new_exp here.

pub fn set_n_dist_bins(&mut self, n_bins: usize) -> &mut Self

Automatically generate distance bins by by using distances from matching sets.

pub fn set_dot_lookup(&mut self, lookup: BinLookup<Precision>) -> &mut Self

Manually set the lookup bins for dot products.

Consider BinLookup::new_linear(0.0, 1.0, n_bins, (true, true)) here.

pub fn set_n_dot_bins(&mut self, n_bins: usize) -> &mut Self

Automatically generate abs dot product bins by using dot products from matching sets.

pub fn set_max_matching_pairs(&mut self, n_pairs: usize) -> &mut Self

Draw at most this many matching pairs from the given matching sets. May be fewer if there aren’t that many combinations. By default, uses all possible pairs.

pub fn set_max_nonmatching_pairs(&mut self, n_pairs: usize) -> &mut Self

Draw at most this many nonmatching pairs from the given nonmatching sets. May be fewer if there aren’t that many combinations. By default, uses as many pairs as for matching.

pub fn build( &self ) -> Result<RangeTable<Precision, Precision>, ScoreMatBuildErr>

Return a tuple of (dist_bins, dot_bins, cells) in the forms accepted by table_to_fn.