Struct goko::plugins::discrete::baseline::DirichletBaseline

pub struct DirichletBaseline { /* fields omitted */ }

Trains a baseline by sampling randomly from the training set (used to create the tree) This baseline is not realistic.

Implementations

impl DirichletBaseline

pub fn set_sequence_len(&mut self, sequence_len: usize)

Sets a new maxium sequence length. Set this to be the window size if you're using windows, the lenght of the test set you've got, or leave it alone as the default limit is the number of points in the training set.

We sample up to this cap, linearly interpolating above this. So, the baseline produced is fairly accurate for indexes below this and unreliable above this.

pub fn set_num_sequences(&mut self, num_sequences: usize)

Sets a new count of sequences to train over, default 100. Stats for each sequence are returned.

pub fn set_prior_weight(&mut self, prior_weight: f64)

Sets a new prior weight, default 1.0. The prior is multiplied by this to increase or decrease it's importance

pub fn set_observation_weight(&mut self, observation_weight: f64)

Sets a new observation weight, default 1.0. Each discrete observation is treated as having this value.

pub fn set_sample_rate(&mut self, sample_rate: usize)

Samples at the following rate, then interpolates for sequence lengths between the following.

pub fn train<D: PointCloud>(
    &self,
    reader: CoverTreeReader<D>
) -> GokoResult<KLDivergenceBaseline>

Trains the sequences up.

Trait Implementations

impl Default for DirichletBaseline

fn default() -> DirichletBaseline

Returns the "default value" for a type.