""" Functions in this file return sampled triplets with answers, based on an artificial embedding and noise.
Usually they combine functions from _triplet_indices and _triplet_answers
and are used as a high-level interface to create artificial datasets.
"""
import numpy as np
from typing import Union
from ._triplet_indices import make_all_triplet_indices
from ._triplet_indices import make_random_triplet_indices
from ._triplet_response import noisy_triplet_response
from .. import utils
def make_all_triplets(embedding: np.ndarray, result_format: str, monotonic: bool = False, **kwargs):
""" Make all possible triplets with answers for the provided embedding or distances.
Args:
embedding: Object coordinates or distance matrix
monotonic: Only triplets (j, i, k), such that j < i < k.
random_state: Seed for noisy answers
kwargs: Additional arguments passed to :func:`cblearn.datasets.noisy_triplet_answers`
Returns:
The triplets and answers, based on format. See :func:`cblearn.utils.check_triplets`.
"""
triplets = make_all_triplet_indices(len(embedding), monotonic)
return noisy_triplet_response(triplets, embedding, result_format=result_format, **kwargs)
[docs]
def make_random_triplets(embedding: np.ndarray, result_format: str, size: Union[int, float] = 1.,
random_state: Union[None, int, np.random.RandomState] = None,
repeat: bool = True, monotonic: bool = False, make_all: int = 10000, **kwargs
) -> utils.Query:
""" Make random triplets with answers for the provided embedding or distances.
>>> triplets, answers = make_random_triplets(np.random.rand(12, 2), size=1000, result_format='list-boolean')
>>> answers.shape, np.unique(answers).tolist()
((1000,), [False, True])
>>> triplets.shape, np.unique(triplets).tolist()
((1000, 3), [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11])
Args:
embedding: Object coordinates (n_objects, n_components) or distance matrix (n_objects, n_objects).
result_format: Result format
size: Either absolute or relative number of triplets to generate.
repeat: Sample triplet indices with repetitions
monotonic: Sample triplets (j, i, k), such that j < i < k.
make_all: Choose from all triplets instead of iterative sampling,
if the difference between all triplets to the requested number is smaller than this value.
random_state: Seed for triplet sampling and noisy answers
kwargs: Additional arguments passed to :func:`cblearn.datasets.noisy_triplet_answers`
Returns:
The triplets and answers, based on format. See :func:`cblearn.utils.check_triplets`.
"""
triplets = make_random_triplet_indices(len(embedding), size, random_state, repeat, monotonic, make_all)
return noisy_triplet_response(triplets, embedding, result_format=result_format, random_state=random_state, **kwargs)
