Pipeline-oriented feature extraction for multimodal datasets
Project description
adfluo
adfluo, adfluis, adfluere, adfluxi, adfluxum
- to flow on/to/towards/by
- to glide/drift quietly
adfluo
is a Python library for pipeline-oriented feature computation, mainly aimed at tricky
multimodal datasets that might require a wide range of different features to be computed from.
Adfluo makes your feature extraction code:
- clean : it encourages you to outline clearly the steps needed to compute a feature as a pipeline of atomic steps
- data scientist-friendly:
adfluo
's output has a predictable structure, ensuring that once you've run the feature extraction, you'll be able to focus a 100% on your data-science/statistics work. - efficient : if different features have computation steps in common,
adfluo
will do its best to compute only what is necessary, without any extra configuration. - reusable: By separating the input data from the feature computation logic, you'll easily be able to reuse an existing extraction pipeline on another dataset, or use another extraction pipeline on the same dataset.
- sample-oriented:
adfluo
organizes its processing around samples of data.
Installation
Adfluo is available on Pypi, and has no special dependencies, a simple
pip install adfluo
will do.
Example
import random
# Defining our dataset as a list of dicts
my_dataset = [
{"numbers" : [random.randint(1, 20) for j in range(50)],
"idx": i}
for i in range(20)
]
# TODO: examples
# - mean, std dev of numbers
# - "relative" mean using idx
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
adfluo-0.3.1.tar.gz
(47.0 kB
view hashes)
Built Distribution
adfluo-0.3.1-py3-none-any.whl
(32.4 kB
view hashes)