Configuration

All visual and algorithmic parameters live in VisualizationConfig, a nested dataclass. The default theme is DARK_SCIENTIFIC.

Structure

VisualizationConfig
├── colors      : ColorPalette     — class colours, background, grid
├── render      : RenderConfig     — canvas size, opacity, marker size
└── projection  : ProjectionConfig — method, scale_input, t-SNE / UMAP params

Fluent helpers

The DARK_SCIENTIFIC singleton provides helpers that return a modified copy:

from geolatent import DARK_SCIENTIFIC

cfg = (
    DARK_SCIENTIFIC
    .with_method("tsne")
    .with_title("My Figure")
    .with_resolution(1400, 900)
    .with_opacity(surface=0.3, scatter=0.9)
)

Manual configuration

For full control, construct or copy the config and mutate its fields:

cfg = DARK_SCIENTIFIC.copy()

# Projection
cfg.projection.method             = "sensitivity"
cfg.projection.scale_input        = True
cfg.projection.sensitivity_n_subsample = 500
cfg.projection.tsne_perplexity    = 40.0
cfg.projection.tsne_n_iter        = 1000

# Rendering
cfg.render.width        = 1200
cfg.render.height       = 800
cfg.render.surface_opacity = 0.25
cfg.render.scatter_opacity = 0.95
cfg.render.marker_size  = 4

Custom colour palette

from geolatent.config.themes import ColorPalette, VisualizationConfig, RenderConfig, ProjectionConfig

my_colors = ColorPalette(
    background="#0a0a0a",
    grid="#1a1a2e",
    text="#ffffff",
    axis_line="#333366",
    class_colors=["#e63946", "#457b9d", "#2a9d8f", "#e9c46a"],
    annotation_color="#8888aa",
)

cfg = VisualizationConfig(
    colors=my_colors,
    render=RenderConfig(width=1100, height=750),
    projection=ProjectionConfig(method="pca"),
)

ProjectionConfig reference

Field	Type	Default	Description
method	str	"pca"	"pca" · "tsne" · "umap" · "sensitivity"
scale_input	bool	True	Standardise features before projection
n_components	int	3	Number of output dimensions
tsne_perplexity	float	30.0	t-SNE perplexity
tsne_n_iter	int	1000	t-SNE iterations
umap_n_neighbors	int	15	UMAP neighbourhood size
umap_min_dist	float	0.1	UMAP minimum distance
sensitivity_n_subsample	int	300	Points used for Jacobian estimation
sensitivity_eps	float	0.01	Finite-difference perturbation magnitude