1 VinPrediction

vin.types.VinPrediction(
    logits,
    prob,
    expected,
    expected_normalized,
    candidate_valid,
    voxel_valid_frac=None,
    semidense_candidate_vis_frac=None,
    semidense_valid_frac=None,
)

VIN predictions for a candidate set.

This is the primary output of aria_nbv.vin.model_v3.VinModelV3. It is consumed by the Lightning training loop (loss + metrics) and by downstream NBV selection (ranking candidates by predicted improvement). The expected score is a learned one-step ranking proxy; rollout-level endpoint metrics and cumulative target RRI are produced by rollout stores and oracle re-scoring, not by this container.

Typical usage in training (see aria_nbv/lightning/lit_module.py): - logits / prob: CORAL ordinal loss and optional auxiliary losses. - expected_normalized: correlation/top-k metrics and candidate ranking proxy. - voxel_valid_frac / semidense_candidate_vis_frac: optional scheduled coverage reweighting of the loss + diagnostics. - candidate_valid: conservative validity heuristic used for logging and optional filtering in analysis/visualization.

1.1 Attributes

Name	Description
logits	Tensor["B N K-1", float32] CORAL logits (K ordinal classes).
prob	Tensor["B N K", float32] Class probabilities derived from CORAL logits.
expected	Tensor["B N", float32] Expected class value in [0, K-1].
expected_normalized	Tensor["B N", float32] Expected value normalized to [0, 1].
candidate_valid	Tensor["B N", bool] Candidate validity mask.
voxel_valid_frac	Tensor["B N", float32] Per-candidate voxel coverage proxy (if available).
semidense_candidate_vis_frac	Tensor["B N", float32] Per-candidate semidense visibility proxy (if available).
semidense_valid_frac	Deprecated alias for semidense_candidate_vis_frac.