Skip to contents

Inverse-distance-weighted active-learning acquisition function using an ALDistance.

The basic score is an IDW-weighted squared residual plus the IDEAL exploration term. With omega > 0, $fit_pool() precomputes the IDEAL pool-density multiplier on the fitted candidate pool. When fitted with xdt = NULL, the population-case density multiplier is set to 1.

Creates a new distance-aware IDEAL acquisition function.

Clears precomputed density values.

Arguments

surrogate

(NULL | mlr3mbo::SurrogateLearner)
Surrogate used for mean predictions.

al_distance

(ALDistance | NULL)
Distance used for the IDW geometry.

delta

(numeric(1))
Exploration weight.

omega

(numeric(1))
Density multiplier weight.

tolerance_equality

(numeric(1))
Squared-distance tolerance for the exact-match branch.

Value

self.