Convert to visualizer — as_visualizer • vistool

This function converts to a visualizer. Automatically chooses between 1D and 2D visualizations based on the number of features/dimensions using ggplot2 backend. For 2D inputs, you can optionally use `type = "surface"` to get interactive plotly surface plots (available for Models and Objectives only).

Usage

as_visualizer(
  x,
  type = "auto",
  x1_limits = NULL,
  x2_limits = NULL,
  padding = 0,
  n_points = 100L,
  y_pred = NULL,
  y_true = NULL,
  input_type = "auto",
  y_curves = "both",
  learner = NULL,
  ...
)

# S3 method for class 'Task'
as_visualizer(
  x,
  type = "auto",
  x1_limits = NULL,
  x2_limits = NULL,
  padding = 0,
  n_points = 100L,
  y_pred = NULL,
  y_true = NULL,
  input_type = "auto",
  y_curves = "both",
  learner = NULL,
  hypothesis = NULL,
  retrain = TRUE,
  ...
)

# S3 method for class 'Hypothesis'
as_visualizer(
  x,
  type = "auto",
  x1_limits = NULL,
  x2_limits = NULL,
  padding = 0,
  n_points = 100L,
  y_pred = NULL,
  y_true = NULL,
  input_type = "auto",
  y_curves = "both",
  learner = NULL,
  domain = NULL,
  ...
)

# S3 method for class 'Objective'
as_visualizer(
  x,
  type = "auto",
  x1_limits = NULL,
  x2_limits = NULL,
  padding = 0,
  n_points = 100L,
  y_pred = NULL,
  y_true = NULL,
  input_type = "auto",
  y_curves = "both",
  learner = NULL,
  allow_extrapolation = FALSE,
  ...
)

# S3 method for class 'LossFunction'
as_visualizer(
  x,
  type = "auto",
  x1_limits = NULL,
  x2_limits = NULL,
  padding = 0,
  n_points = 1000L,
  y_pred = NULL,
  y_true = NULL,
  input_type = "auto",
  y_curves = "both",
  learner = NULL,
  ...
)

# S3 method for class 'list'
as_visualizer(
  x,
  type = "auto",
  x1_limits = NULL,
  x2_limits = NULL,
  padding = 0,
  n_points = 1000L,
  y_pred = NULL,
  y_true = NULL,
  input_type = "auto",
  y_curves = "both",
  learner = NULL,
  ...
)

Arguments

x: (`any`)
Object to convert to a visualizer.
type: (`character(1)`)
The type of visualization: "auto" (default), "1d", "2d", or "surface". If "auto", automatically chooses between 1D and 2D ggplot2 visualizations based on the number of features/dimensions. Use "surface" for interactive plotly surface plots (2D inputs only, Models and Objectives only).
x1_limits: (`numeric(2)`)
The x1 limits.
x2_limits: (`numeric(2)`)
The x2 limits.
padding: (`numeric(1)`)
A margin that is added to x1limits and x2limits. The x1 margin is calculated by `max(x1lmits) - min(x1limits) * padding`.
n_points: (`integer(1)`)
The number of generated point per dimension. Note that a grid of `npoints^2` values is generated and evaluated by `objective$eval(x)` to plot the surface.
y_pred: (`numeric()`)
Predicted values (used for loss function visualizations).
y_true: (`numeric()`)
True values (used for loss function visualizations).
input_type: (`character(1)`)
`"auto"` (default), `"score"` or `"probability"`. Passed through to the loss visualiser.
y_curves: (`character(1)`)
Which response curve(s) to draw when `input_type = "probability"`. One of `"both"`, `"y1"`, or `"y0"`.
learner: (`mlr3::Learner`)
The learner to train the model with.
...: (`any`)
Additional arguments.
hypothesis: (`Hypothesis`|`NULL`) Optional hypothesis object that provides predictions instead of a learner. Supply either `learner` or `hypothesis`, but not both.
retrain: (`logical(1)`)
If a learner is supplied, whether it should be (re)trained on the task (`TRUE`, default) or reused as-is (`FALSE`). If `FALSE` but the learner is untrained, a warning is issued and training is performed.
domain: (`list`|`NULL`) Named list giving axis limits per predictor when no Task data is available, e.g., `list(x = c(-1, 1))` for 1D or `list(x = c(-1,1), y = c(-1,1))` for 2D.
allow_extrapolation: (`logical(1)`) Whether to evaluate the objective outside its evaluation (canonical) bounds when plot limits extend further. If `FALSE` (default), values outside are masked (set to `NA`) producing blank regions in the plot; if `TRUE`, the objective is called on the extended region (may yield misleading or numerically extreme values depending on the function definition).

Value

An object inheriting from a Visualizer class (Visualizer1D, Visualizer2D, VisualizerSurface, etc.) depending on the input and selected type.

Details

If `type = "auto"` (default), the function will inspect the input and select the appropriate ggplot2 visualizer: - 1D: For objects with 1 feature/dimension (uses ggplot2) - 2D: For objects with 2 features/dimensions (uses ggplot2) You can override this by specifying `type = "1d"`, `type = "2d"`, or for 2D inputs only: `type = "surface"` (uses plotly for interactive surfaces, Models and Objectives only).

Hypotheses can be used instead of learners to visualize functional forms directly. When no Task is provided (i.e., visualizing a `Hypothesis` alone), a `domain` must be supplied to define plotting limits.