UncertainPrediction

class UncertainPrediction(score: FloatTensor, uncertainty: FloatTensor)

Bases: NamedTuple

A pair of predicted scores and corresponding uncertainty.

Since the uncertainty scores come from Monte-Carlo dropout, they are guaranteed to be non-negative with larger scores indicating higher uncertainty.

Create new instance of UncertainPrediction(score, uncertainty)

Attributes Summary

score	The scores
uncertainty	The uncertainty, in the same shape as scores

Methods Summary

from_scores(scores)

Make an instance from scores.

Attributes Documentation

Parameters:

• score (FloatTensor) –

• uncertainty (FloatTensor) –

score: FloatTensor

The scores

uncertainty: FloatTensor

The uncertainty, in the same shape as scores

Methods Documentation

classmethod from_scores(scores)

Make an instance from scores.

Parameters:

scores (Tensor) –