Predictive distributions - Prior Labs

Looking for usage documentation? Check out Predictive distribution and Regression.

`visualisation.plot_regression_distribution`

Plot the predicted target distribution for a single sample.

visualisation.plot_regression_distribution(
    prediction: Mapping[str, Any],
    *,
    sample_idx: int = 0,
    statistics: Sequence[str] = ("mean", "median", "mode"),
    quantile_interval: tuple[float, float] | None = (0.1, 0.9),
    zoom_quantile: float | None = 0.99,
    smooth: float = 0.005,
    ax: Axes | None = None,
    color: str = "#1f77b4",
) -> Axes

Parameters

Parameter	Type	Default	Description
`prediction`	`Mapping[str, Any]`	Required	Output of `regressor.predict(X, output_type="full")`. It may hold several samples; pick the one to plot with `sample_idx`.
`sample_idx`	`int`	`0`	Index of the sample to plot within `prediction`.
`statistics`	`Sequence[str]`	`("mean", "median", "mode")`	Point statistics to mark with a vertical line. Each one must be present in `prediction`.
`quantile_interval`	`tuple[float, float] \| None`	`(0.1, 0.9)`	Central interval to shade. Pass `None` to disable.
`zoom_quantile`	`float \| None`	`0.99`	Fraction of probability mass to keep in view, centred on the median. Pass `None` to show the full support.
`smooth`	`float`	`0.005`	Width of the display-only moving average over the density, as a fraction of the number of bars. Pass `0` to show the raw bar density.
`ax`	`Axes \| None`	`None`	Existing axes to draw on. A new figure is created if omitted. When the axes already holds a curve, the limits, labels and legend of that curve are preserved so several distributions can be overlaid.
`color`	`str`	`"#1f77b4"`	Base colour of the density curve.

Returns

Type	Description
`Axes`	The axes containing the plot.

Raises ValueError If prediction or the other arguments are invalid. ModuleNotFoundError If matplotlib is not installed. Plotting requires pip install "tabpfn-client[viz]".