spartans package

Top-level package for Spartans.

Submodules

spartans.core module

spartans.core.asarray(a)[source]

convenience - turn np.matrix to np.array including dim reduction

spartans.core.constant_index(x, axis, mask=None, as_bool=True, threshold=0)[source]

Returns the indices of the constant rows/features

x : sparse matrix
data matrix
axis : int
axis to return indices for
mask : sparse matrix [bool]
mask of values to consider as nan
as_bool : bool
whether to return a mask of bool indices or vector of numbers
threshold : numeric
decided constant by the feature variance, can be larger than 0 for “almost-constant” features
cond : array_like
Either an array with number of indices or boolean mask
spartans.core.corr(x, y=None, mask=None)[source]

Return a correlation vector between matrix x and target column y if given, else a auto-correlation matrix for the features of matrix x

x : sparse matrix
Data Matrix
y : array_like
target array
mask : sparse matrix [bool]
mask of values to consider as nan
ret
correlation vector if y in given, also auto-correlation for x
spartans.core.cov(x, y=None, mask=None, blocks=1)[source]
x : sparse matrix
Data Matrix
y : array_like
target array
mask : sparse matrix [bool]
mask of values to consider as nan
blocks : int
amount of blocks of computing (for large matrices)
ret
covariance vector if y in given, also auto-covariance for x
spartans.core.density(x)[source]
spartans.core.make_nan_mask(x)[source]
spartans.core.mean(x, axis=None, mask=None, safe=False, **kwargs)[source]
spartans.core.non_constant_index(x, axis, mask=None, as_bool=True, threshold=0, method='variance')[source]

Returns the indices of the non constant (informative) rows/features

x : sparse matrix
data matrix
axis : int
axis to return indices for
mask : sparse matrix [bool]
mask of values to consider as nan
as_bool : bool
whether to return a mask of bool indices or vector of numbers
threshold : numeric
decided constant by the feature variance, can be larger than 0 for “almost-constant” features
cond : array_like
Either an array with number of indices or boolean mask
spartans.core.non_zero_index(x, axis, as_bool=True)[source]

return the index of all rows/features that are not all zero

x : sparse matrix
data matrix
axis : int
axis to return indices for
mask : sparse matrix [bool]
mask of values to consider as nan
as_bool : bool
whether to return a mask of bool indices or vector of numbers
cond : array_like
Either an array with number of indices or boolean mask
spartans.core.variance(x, axis=None, mask=None, **kwargs)[source]

Returns variance by axis or for entire sparse matrix

mask x : sparse.csr_matrix

matrix to compute variance for
axis : int or None
axis to return variance for, or None if for entire matrix
kwargs
passed to np.mean
var_ : array_like
array of ndim=1 if axis is given or 0 dim (scalar) if axis is None