Skip to contents

Subspace task definition

TaskSubspace.Rd

R6 class for defining hyperparameter optimization tasks. Encapsulates benchmark data with hyperparameter configurations and performance measures, providing a standardized interface for subspace learners.

Details

Overview:

A subspace task combines:

  • Benchmark data with hyperparameter configurations and performance measures

  • Specification of which columns are hyperparameters (continuous and categorical)

  • Target performance measure to optimize

Data Requirements:

The input data.table must contain:

  • A task column identifying different tasks/datasets

  • A target measure column (e.g., "auc", "accuracy", "rmse")

  • One or more numeric hyperparameter columns

  • Optionally, one categorical hyperparameter column

Initialization Modes:

Mode 1: Explicit specification


task <- TaskSubspace$new(
  data = benchmark_data,
  target_measure = "accuracy",
  hps = c("learning_rate", "max_depth"),
  cat_hps = "optimizer"
)

Mode 2: Formula interface


task <- TaskSubspace$new(
  data = benchmark_data,
  formula = accuracy ~ (learning_rate + max_depth) * optimizer
)

The formula syntax is: target ~ (hp1 + hp2 + ...) * cat_hp

  • Left-hand side: target measure

  • Right-hand side before *: continuous hyperparameters

  • Right-hand side after *: categorical hyperparameter (optional)

Categorical Hyperparameters:

Currently supports at most one categorical hyperparameter. When specified, learners will fit separate subspaces for each categorical level.

See also

LearnerSubspace for the base learner class. LearnerSubspaceBox for axis-aligned hyperrectangles. LearnerSubspacePolygon for oriented hyperrectangles. LearnerSubspaceEllipsoid for ellipsoids.

Public fields

data

A data.table containing hyperparameter configurations, performance measures, and task identifiers. Must have a task column.

target_measure

Character string specifying the performance measure column name (e.g., "auc", "accuracy", "rmse")

hps

Character vector of continuous hyperparameter column names

cat_hps

Character string specifying the categorical hyperparameter column name (optional, currently limited to one)

Methods

Public methods

Method new()

Create a new task instance

Usage

TaskSubspace$new(
  data,
  formula = NULL,
  target_measure = NULL,
  hps = NULL,
  cat_hps = NULL
)

Arguments

data

A data.table containing task performance data

formula

Formula specification (optional)

target_measure

Name of the performance measure column

hps

Character vector of continuous hyperparameter names

cat_hps

Character vector of categorical hyperparameter names (optional)

Method clone()

The objects of this class are cloneable with this method.

Usage

TaskSubspace$clone(deep = FALSE)

Arguments

deep

Whether to make a deep clone.

Examples

if (FALSE) { # \dontrun{
# Explicit specification
task <- TaskSubspace$new(
  data = benchmark_data,
  target_measure = "auc",
  hps = c("learning_rate", "max_depth", "min_samples_split")
)

# With categorical hyperparameter
task <- TaskSubspace$new(
  data = benchmark_data,
  target_measure = "accuracy",
  hps = c("learning_rate", "max_depth"),
  cat_hps = "optimizer"
)

# Formula interface
task <- TaskSubspace$new(
  data = benchmark_data,
  formula = auc ~ (learning_rate + max_depth) * optimizer
)

# Use with learner
learner <- LearnerSubspaceBox$new(task)
learner$train(q_val = 0.9)
} # }