spartans 0.2.0

Sparse Tools for Analysis

spartans - SPARse Tools for ANalysiS

When working with sparse matrices, it is desired to have a way to work with them as if they were a regular numpy.arrays. Yet, many popular methods for arrays don’t exist for sparse matrices. spartans wishes to help, with many operations to work with

Full example notebook

• Free software: GNU General Public License v3

• Documentation: https://spartans.readthedocs.io.

Features

Mathematical Operations

Rich set of operations not supported on sparse matrices like variance, cov (covariance matrix) and corrcoef (correlation matrix).

Easy Indexing

Convenient methods to index for “extra” sparse features by variance or by quantity.

Masking

Many algorithms consider the zeros in a sparse matrix as missing data. Or considering missing data as zeros. Depending on the use-case. spartans

FeatureMatrix

FeatureMatrix is a spartan's first-class citizen. It is a wrapper around scipy.sparse.csr Matrix built with data analysis and data-science in mind.

Examples

Full example notebook

>>> import spartans as st
>>> from scipy.sparse import csr_matrix
>>> import numpy as np
>>> m = np.array([[1, -2, 0, 50],
                  [0, 0, 0, 100],
                  [1, 0, 0, 80],
                  [1, 4, 0, 0],f
                  [0, 0, 0, 0],
                  [0, 4, 0, 0],
                  [0, 0, 0, -50]])
>>> c = csr_matrix(m)

We can get the the correlation matrix of m using numpy.

>>> np.corrcoef(m, rowvar=False)

Out[]: array([[ 1.  , -0.08,   nan,  0.31],
              [-0.08,  1.  ,   nan, -0.35],
              [  nan,   nan,   nan,   nan],
              [ 0.31, -0.35,   nan,  1.  ]])

This won’t work with the sparse matrix c

>>> np.corrcoef(c, rowvar=False)

AttributeError: 'float' object has no attribute 'shape'

But with spartans this can be done.

>>> st.corr(c)

Out[]: array([[ 1.  , -0.08,   nan,  0.31],
              [-0.08,  1.  ,   nan, -0.35],
              [  nan,   nan,   nan,   nan],
              [ 0.31, -0.35,   nan,  1.  ]])

The column and row with nan is because the original matrix has a columns (feature) which is zero for the entire column. spartans can handle that using st.non_zero_index(c, axis=0, as_bool=False) which will return array([0, 1, 3]). A lot more functionality is in the notebook.

Credits

• This open-source project is backed by SentinelOne

• This package was created with Cookiecutter and the audreyr/cookiecutter-pypackage project template.

History

0.1.0 (2020-02-20)

• First release on PyPI.