PCAScanResult

class PCAScanResult(x_offsets, y_offsets, values, parameter_sets, center, principal_component_x, principal_component_y, explained_variance_ratio, projected_samples, scan_component_ids, program_type)

Bases: object

Attributes Summary

center
explained_variance_ratio
parameter_sets
principal_component_x
principal_component_y
program_type
projected_samples
scan_component_ids
values
x_offsets
y_offsets

Methods Summary

plot(*[, ax, show, levels, add_colorbar, ...])	Plot the sampled PCA landscape and return (fig, ax).
plot_3d(*[, ax, show])	Render the PCA landscape as a 3D surface and return (fig, ax).

Attributes Documentation

center: ndarray[tuple[Any, ...], dtype[float64]]

explained_variance_ratio: ndarray[tuple[Any, ...], dtype[float64]]

parameter_sets: ndarray[tuple[Any, ...], dtype[float64]]

principal_component_x: ndarray[tuple[Any, ...], dtype[float64]]

principal_component_y: ndarray[tuple[Any, ...], dtype[float64]]

program_type: str

projected_samples: ndarray[tuple[Any, ...], dtype[float64]]

scan_component_ids: tuple[int, int]

values: ndarray[tuple[Any, ...], dtype[float64]]

x_offsets: ndarray[tuple[Any, ...], dtype[float64]]

y_offsets: ndarray[tuple[Any, ...], dtype[float64]]

Methods Documentation

plot(*, ax=None, show=False, levels=20, add_colorbar=True, show_samples=True, sample_kwargs=None, show_trajectory=False, trajectory_kwargs=None, show_gradients=False, gradient_kwargs=None, **contour_kwargs)

Plot the sampled PCA landscape and return (fig, ax).

Parameters:

• ax – Optional matplotlib axes to draw into. When omitted, a new figure and axes are created.

• show (bool) – Whether to call matplotlib.pyplot.show() after drawing.

• levels (int) – Number of discrete color bands (BoundaryNorm) for pcolormesh.

• add_colorbar (bool) – Whether to attach a colorbar to the figure.

• show_samples (bool) – Whether to overlay the projected PCA samples on top of the heatmap.

• sample_kwargs (dict | None) – Optional keyword arguments forwarded to ax.scatter for the PCA sample overlay.

• show_trajectory (bool) – Whether to draw a connected line through projected_samples in order. The samples must have been supplied in temporal order (e.g. from param_history()).

• trajectory_kwargs (dict | None) – Optional keyword arguments forwarded to ax.plot for the trajectory line. Defaults to a thin white line with start/end markers.

• show_gradients (bool) – Whether to overlay a quiver plot of the numerical gradient (computed via numpy.gradient() on the grid).

• gradient_kwargs (dict | None) – Optional keyword arguments forwarded to ax.quiver for the gradient overlay.

• **contour_kwargs – Additional keyword arguments forwarded to ax.pcolormesh (corner_mask is ignored). PCA scans use a cell heatmap instead of contourf so noisy objectives still fill the axes.

Returns:

(fig, ax) for the rendered plot.

Return type:

tuple

plot_3d(*, ax=None, show=False, **surface_kwargs)

Render the PCA landscape as a 3D surface and return (fig, ax).

Parameters:

• ax – Optional Axes3D. When omitted, a new figure with a 3D projection is created.

• show (bool) – Whether to call matplotlib.pyplot.show() after drawing.

• **surface_kwargs – Additional keyword arguments forwarded to ax.plot_surface.

Returns:

(fig, ax) for the rendered 3D plot.

Return type:

tuple