semsynth.umap_utils

Utilities for building and rendering UMAP embeddings lazily.

Functions

`build_umap`(df, discrete_cols, ...[, ...])	Fit a UMAP model on a sample of the dataset.
`dataclass`([cls, init, repr, eq, order, ...])	Add dunder methods based on the fields defined in the class.
`pick_color_labels`(series)	Convert a categorical series into numeric labels for coloring.
`plot_umap`(embedding, outfile, title[, ...])	Render the embedding to disk and return axis limits.
`transform_with_umap`(art, df)	Project new data with the fitted UMAP model.

Classes

`Any`(args, *kwargs)	Special type indicating an unconstrained type.
`Path`(args, *kwargs)	PurePath subclass that can make system calls.
`UMAPArtifacts`(preproc, umap_model, ...)	Artifacts generated when fitting a UMAP projection.

class semsynth.umap_utils.UMAPArtifacts(preproc: Any, umap_model: Any, sample_idx: np.ndarray, embedding: np.ndarray, label_mapping: Dict[Any, int] | None, color_labels: 'np.ndarray' | None)

Bases: object

Artifacts generated when fitting a UMAP projection.

color_labels: 'np.ndarray' | None

embedding: np.ndarray

label_mapping: Dict[Any, int] | None

preproc: Any

sample_idx: np.ndarray

umap_model: Any

semsynth.umap_utils.build_umap(df: pd.DataFrame, discrete_cols: List[str], continuous_cols: List[str], color_series: 'pd.Series' | None, rng: np.random.Generator, random_state: int = 42, max_sample: int = 1000, n_neighbors: int = 30, min_dist: float = 0.1, n_components: int = 2) → UMAPArtifacts: Fit a UMAP model on a sample of the dataset.

semsynth.umap_utils.pick_color_labels(series: 'pd.Series' | None) → Tuple['np.ndarray' | None, Dict[Any, int] | None]: Convert a categorical series into numeric labels for coloring.

semsynth.umap_utils.plot_umap(embedding: np.ndarray, outfile: str, title: str, color_labels: 'np.ndarray' | None = None, lims: Tuple[Tuple[float, float], Tuple[float, float]] | None = None) → Tuple[Tuple[float, float], Tuple[float, float]]: Render the embedding to disk and return axis limits.

semsynth.umap_utils.transform_with_umap(art: UMAPArtifacts, df: pd.DataFrame) → np.ndarray: Project new data with the fitted UMAP model.