Skip to contents

Hypothesis wrapper for functional models

Hypothesis.Rd

Construct a hypothesis object that wraps a user-supplied function with minimal metadata so it can be visualized like a learner.

R6 class wrapping a user-supplied function plus minimal metadata so it can be visualized like a learner on 1D or 2D inputs.

Usage

hypothesis(
  fun,
  type,
  predictors,
  link = "identity",
  domain = NULL,
  levels = NULL
)

Arguments

fun

(`function`) Function defining the hypothesis. Accepted signatures: - 1D: `fun(x, ...)` or `fun(data)` where `data` is a data.frame with one predictor column - 2D: `fun(x, y, ...)` or `fun(data)` where `data` is a data.frame with two predictor columns

type

(`character(1)`) One of "regr" or "classif".

predictors

(`character()`) Names of the predictor columns. Length must be 1 or 2.

(`character(1)`) Link for classif outputs: "identity" (default) or "logit". For regression, only "identity" is used.

domain

(`list`|`NULL`) Named list with limits per predictor, e.g., `list(x = c(-3, 3))` for 1D or `list(x = c(-3,3), y = c(-3,3))` for 2D. Used when no Task is provided.

levels

(`character(2)`|`NULL`) Class labels for binary classification.

Value

An `R6` Hypothesis object with a `predict(newdata)` method.

Public fields

fun

(`function`) User-provided function implementing the hypothesis.

type

(`character(1)`) Prediction type, one of "regr" or "classif".

predictors

(`character()`) Names of predictor columns (length 1 or 2).

input_dim

(`integer(1)`) Number of predictors, derived from `predictors`.

link

(`character(1)`) Link for classif output: "identity" or "logit".

domain

(`list`|`NULL`) Named list with limits per predictor for plotting without a Task.

levels

(`character(2)`|`NULL`) Class labels for binary classification.

Methods

Public methods

Method new()

Create a new Hypothesis instance

Usage

Hypothesis$new(
  fun,
  type,
  predictors,
  link = "identity",
  domain = NULL,
  levels = NULL
)

Arguments

fun

(`function`) See class field `fun`.

type

(`character(1)`) See class field `type`.

predictors

(`character()`) See class field `predictors`.

link

(`character(1)`) See class field `link`.

domain

(`list`|`NULL`) See class field `domain`.

levels

(`character(2)`|`NULL`) See class field `levels`.

Method predict()

Predict on newdata

Usage

Hypothesis$predict(newdata)

Arguments

newdata

(`data.frame`) with columns matching `predictors` in order

Returns

numeric vector of predictions (regression) or probabilities in [0,1] (classification)

Method clone()

The objects of this class are cloneable with this method.

Usage

Hypothesis$clone(deep = FALSE)

Arguments

deep

Whether to make a deep clone.