Struct Tpe 

pub struct Tpe {
    pub goal: Goal,
    pub startup: usize,
    pub candidates: usize,
    pub quantile: f64,
    pub seed: u64,
}

Tree-structured Parzen Estimator: model what the good trials did, model what the bad ones did, and propose where the first is likely and the second is not.

The one that looks at what already happened, which is also its one honest cost: it cannot derive trial 7 from the seed and the index alone, so a study spread over a folder gets a different search than one in a single process. That is what being guided means, not a bug to fix.

Fields

goal: Goal

Which way is better.

startup: usize

Draw at random until this many trials have finished. Below two there is nothing to split into good and bad, so two is the floor whatever is said.

candidates: usize

How many places to consider before proposing one. More is a better proposal and costs nothing but arithmetic — no trial is run for it.

quantile: f64

What share of the finished trials counts as good. 0.25 keeps the best quarter as the thing to imitate.

seed: u64

The seed of the draws.

Implementations

impl Tpe

pub fn ask( &self, space: &Space, trial: usize, seen: &[(Point, Option<f64>)], ) -> Option<Point>

The trial-th point. Random until startup, guided after.

Trait Implementations

impl Clone for Tpe

fn clone(&self) -> Tpe

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for Tpe

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

Formats the value using the given formatter.

impl From<Tpe> for Sampler

fn from(how: Tpe) -> Self

Converts to this type from the input type.

impl PartialEq for Tpe

fn eq(&self, other: &Tpe) -> bool

Tests for self and other values to be equal, and is used by ==.

fn ne(&self, other: &Rhs) -> bool

Tests for !=. The default implementation is almost always sufficient, and should not be overridden without very good reason.

impl StructuralPartialEq for Tpe