scores.RandomScore

Returns random uncertainty scores per sample.

Usage

Source

scores.RandomScore()

This score assigns a random float in [0, 1] to each sample. It is useful as a baseline or for testing purposes. By default, the threshold is set to 0.99, so approximately 99% of samples are selected.

Methods

Name	Description
fit()	Unused.
score()	Compute a random uncertainty score for every sample in a batch.
select()	Select samples for prediction based on their random uncertainty score.

---

fit()

Unused.

Usage

Source

fit(X=None, Y=None)

---

score()

Compute a random uncertainty score for every sample in a batch.

Usage

Source

score(X)

Returns random scores where low values indicate likely inliers and high values indicate likely outliers.

Parameters

X: torch.Tensor

Input batch of shape (B, ...). Only the batch size is used.

Returns

torch.Tensor

1-D tensor of shape (B,) with uniform random scores in [0, 1].

Examples

import torch
from seapig.scores import RandomScore
score = RandomScore()
scores = score.score(torch.zeros(4, 10))

---

select()

Select samples for prediction based on their random uncertainty score.

Usage

Source

select(X)

Samples with scores lower than the threshold are selected for prediction.

Parameters

X: torch.Tensor

Input batch of shape (B, ...). Only the batch size is used.

Returns

dict[str, torch.Tensor]

A dict with keys 'score' (random scores) and 'selected' (boolean mask where True means the sample is selected).

Examples

import torch
from seapig.scores import RandomScore
score = RandomScore()
result = score.select(torch.zeros(4, 10))
# result['selected'] is a boolean tensor of shape (4,)

See Also

scores.UncertaintyScore