Skip to main content

Python tools to sample randomly with dont pick closest `n` elements constraints. Also contains a batch generator for the same to sample with replacement and with repeats if necessary.

Project description

Sampling Utils

Pypi Version Pypi Licence Pypi Wheel

Python tools to sample randomly with dont pick closest n elements constraints. Also contains a batch generator for the same to sample with replacement and with repeats if necessary.


Simply install using pip

pip install sampling_utils


Dont Pick Closest

from sampling_utils import sample_from_list
sample_from_list([1,2,3,4,5,6,7,8], dont_pick_closest=2)

You are guaranteed to get samples that are at least dont_pick_closest apart# (in value, not in indices). Here you will get samples where sample - any_other_sample is always greater than 2.

For example, if 2 is picked, no other item in range [2+dont_pick_closest and 2-dont_pick_closest] will be picked

Another example looped 5 times:

for _ in range(5):
    sample_from_list([1,2,3,4,5,6,8,9,10,12,14], dont_pick_closest=2)

# Output
# [5, 10, 2, 14]
# [9, 6, 14, 1]
# [3, 8, 12]
# [10, 3, 6, 14]
# [2, 5, 8, 12]

If 12 is sampled, sampling 10 and 14 are not allowed since dont_pick_closest is 2. In other words, if n is sampled, then sampling anything from [n-dont_pick_closest, ... n-1, n , n+1, ... n+dont_pick_closest] is not allowed (if present in the list).

#Will be called as dont_pick_closest rule hereafter.

Number of samples

You can also specify how many samples you want from the list using number_of_samples parameter. By default, you get maximum possible samples (without replacement).

for _ in range(5):
    sample_from_list([1,2,3,4,5,6,8,9,10,12,14], dont_pick_closest=2, num_samples=2)

# Output
# [8, 2]
# [6, 3]
# [12, 1]
# [4, 10]
# [9, 1]

If you try to sample more than what's possible, you will get an error saying that it's not possible.

Min and max samples

You may want to just know how much you can sample from a given list obeying the dont_pick_closest rule

from sampling_utils import get_min_samples, get_max_samples
print(get_min_samples([1,2,3,4,5,6,8,9,10,12,14], dont_pick_closest=2))
print(get_max_samples([1,2,3,4,5,6,8,9,10,12,14], dont_pick_closest=2))

# Output
# Min 3
# Max 4

Sampling without replacement successively / Generating batches of samples for one epoch

If you want to successively sample without replacement i.e. sample as many samples from the list without repeating, you can use batch_rand_generator as shown below. This is particularly useful to generate batches of data until no more batches can be generated (equivalent to one epoch).

from sampling_utils import batch_rand_generator 
from sampling_utils import get_batch_generator_elements

batch_size = 2
brg = batch_rand_generator([1,2,3,4,5,6,8,9,10,12,14], batch_size=batch_size, dont_pick_closest=2)
print(get_batch_generator_elements(brg, batch_size=batch_size))
# Output
# [[1, 4], [8, 5], [14, 3], [2, 6]]

Notice that the elements

  • within each batch obey the dont_pick_closest rule (e.g. 1 and 4 from batch 1)
  • from different batches need not obey the rule (e.g. 4 and 5 from batch 1 and 2 respectively).


Pull requests are very welcome.

  1. Fork the repo
  2. Create new branch with feature name as branch name
  3. Check if things work with a jupyter notebook
  4. Raise a pull request


Please see attached Licence

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Files for sampling-utils, version 0.1.1
Filename, size File type Python version Upload date Hashes
Filename, size sampling_utils-0.1.1-py3-none-any.whl (6.3 kB) File type Wheel Python version py3 Upload date Hashes View
Filename, size sampling_utils-0.1.1.tar.gz (4.9 kB) File type Source Python version None Upload date Hashes View

Supported by

AWS AWS Cloud computing Datadog Datadog Monitoring DigiCert DigiCert EV certificate Facebook / Instagram Facebook / Instagram PSF Sponsor Fastly Fastly CDN Google Google Object Storage and Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Salesforce Salesforce PSF Sponsor Sentry Sentry Error logging StatusPage StatusPage Status page