MF Algorithms contains various matrix factorization methods utilizing different iterative update rules.
Project description
MF Algorithms
MF Algorithms contains various matrix factorization methods utilizing different iterative update rules.
Installation
To install MF Algorithms, run this command in your terminal:
$ pip install -U mf-algorithms
This is the preferred method to install MF Algorithms, as it will always install the most recent stable release.
If you don't have pip installed, these installation instructions can guide you through the process.
Usage
First import functions
from the package. scipy.sparse
is also useful for creating toy sparse matrices to test the algorithms, thought we will manually generate factor matrices and multiply them to guarantee its rank.
>>> import numpy as np
>>> from mf_algorithms import functions
Matrix Factorization
>>> dim1 = 1000
>>> dim2 = 1000
>>> k = 50
>>> factors = np.random.choice(4, size=(dim1,k), p=np.array([0.97, 0.01, 0.01, 0.01]))
>>> weights = np.random.choice(2, size=(k, dim2), p=np.array([0.999, 0.001]))
>>> mat = factors @ weights
>>> A, S, error = functions.mf(data = mat, k = 50, s1 = 1, s2 = 1, niter = 100, siter = 1, update = 'als', errseq = False)
Citing
If you use our work in an academic setting, please cite our paper:
Development
See CONTRIBUTING.md for information related to developing the code.
Suggested Git Branch Strategy
master
is for the most up-to-date development, very rarely should you directly commit to this branch. Your day-to-day work should exist on branches separate frommaster
. It is recommended to commit to development branches and make pull requests to master.4. It is recommended to use "Squash and Merge" commits when committing PR's. It makes each set of changes tomaster
atomic and as a side effect naturally encourages small well defined PR's.
Additional Optional Setup Steps:
-
Create an initial release to test.PyPI and PyPI.
- Follow This PyPA tutorial, starting from the "Generating distribution archives" section.
-
Create a blank github repository (without a README or .gitignore) and push the code to it.
-
Delete these setup instructions from
README.md
when you are finished with them.
Project details
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Hashes for mf_algorithms-0.2.2-py2.py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 6efbd67371aebcf3432496899f1b02fd3ff9690bd486a459f74f3943aa3e1e20 |
|
MD5 | f1cec1bd6edfe22a1ad794931d62b1b6 |
|
BLAKE2b-256 | 028834d4c27322ca11f707f9af9a84b6aa26f1d9eaf039dca635bed7f501ba1e |