Regression

The regression task head maps the aggregated bag representation to a continuous target, trains with MSE, and reports mae and r2. The aggregator is the method home for the bag → slide-level step that feeds the head — see Aggregators.

class soma.tasks.regression.RegressionHead(input_dim, num_targets=1, metrics=None)

Bases: TaskHead

Linear regression head for single or multi-target regression.

Parameters:

• input_dim (int) – Dimension of the input representation.

• num_targets (int) – Number of regression targets. Default 1.

• metrics (list[str] | None) – Metrics to compute. Empty list uses the default set for regression: mae, r2.