torchcell.transforms.coo_regression_to_classification.COOLabelNormalizationTransform

class COOLabelNormalizationTransform(dataset: Neo4jCellDataset, label_configs: dict[str, dict[str, Any]], eps: float = 1e-08, fit_indices: Iterable[int] | None = None)[source]

Bases: BaseTransform

Transform for normalizing labels in COO format with different strategies.

normalize(values: Tensor, label: str) → Tensor[source]: Normalize values based on specified strategy.

denormalize(values: Tensor, label: str) → Tensor[source]: Denormalize values based on specified strategy.

forward(data: HeteroData | Batch) → HeteroData | Batch[source]: Transform the data by normalizing specified labels in COO format.

inverse(data: HeteroData | Batch) → HeteroData | Batch[source]: Inverse transform to recover original scale.