espnet2.layers.augmentation.weighted_sample_without_replacement

source

espnet2.layers.augmentation.weighted_sample_without_replacement(population, weights, k, rng=<module 'random' from '/data/user_data/msomeki/miniconda/lib/python3.10/random.py'>)

Sample k unique elements from a population based on provided weights.

This function samples k elements from a given population without replacement, ensuring that each element is sampled according to its specified weight. It uses a probabilistic method to select elements, which means that elements with higher weights have a greater chance of being selected.

• Parameters:
  • population (list) – A list of elements from which to sample.
  • weights (list) – A list of weights corresponding to each element in the population. The weights must be positive and have the same length as the population.
  • k (int) – The number of unique elements to sample. Must be less than or equal to the length of the population.
  • rng (random.Random , optional) – An optional random number generator. If not provided, the default random module will be used.
• Returns: A list of k unique elements sampled from the population.
• Return type: list
• Raises:ValueError – If k is greater than the length of the population.

Examples

>>> population = ['apple', 'banana', 'cherry', 'date']
>>> weights = [0.1, 0.2, 0.3, 0.4]
>>> weighted_sample_without_replacement(population, weights, 2)
['cherry', 'date']  # Example output may vary due to randomness.

>>> weighted_sample_without_replacement(population, weights, 0)
[]  # Sampling 0 elements returns an empty list.

NOTE

The function ensures that no element is selected more than once.