v0.1.16

What's Changed

Changelog v0.1.16:

• Extend Python version compatibility to 3.11
• Minor code improvements
• Remove unnecessary traces of Multinomial distribution
• Fix issues in not installing the tests package
• Add black to dev requirements
• Reformat the codes using black
• Minor change in docstrings

v0.1.15

What's Changed

Changelog v0.1.15:

• New feature: GeneralNB takes a smoothing parameter alpha for estimating distribution parameters (currently, for Bernoulli and Categorical distributions)
• Reformat the codes using black
• Minor change in docstrings
• Minor change in error messages

v0.1.14

What's Changed

Changelog v0.1.14:

• Remove MultinomialDist from the supported distributions
• Use a constant smoothing parameter (1e-10) in BernoulliDist and CategoricalDist when estimating their PMFs
• Add support for nun-numerical features for some distributions (e.g., categorical distribution)
• Add new tests methods to compare GeneralNB with sklearn's BernoulliNB and CategoricalNB
• Minor bug fixing and improvements
• Minor update in docstrings

v0.1.13

What's Changed

Changelog v0.1.13:

• Add more unit tests for GaussianWNB
• Move learning_hist parameter from fit method to __init__ method
• Now, cost_hist_ attribute will store None if learning_hist=False

v0.1.12

What's Changed

Changelog v0.1.12:

• Add unit tests for GaussianWNB
• Refactor and minor improvements in GaussianWNB

v0.1.11

What's Changed

Changelog v0.1.11:

• Test whether the GeneralNB estimator adheres to scikit-learn conventions
• Convert some hidden methods to internal methods
• Fix classifiers in setup.py
• Add pypi build badge to README

v0.1.10

What's Changed

Changelog v0.1.10:

• np.mean() is replaced with np.average() to estimate the expected value of normal distribution
• Finding classes and their count is done first in the fit method
• Minor change in error messages

v0.1.9

What's Changed

Changelog v0.1.9:

• Rename priors_ to class_prior_ in GeneralNB and GaussianWNB classes
• Remove ddof=1 while calculating standard deviation for normal distribution
• Fix a critical bug in calculating log probability in GeneralNB class
• Add a tests directory with some unit tests for GeneralNB

v0.1.8

What's Changed

Changelog v0.1.8:

• Add _support attribute and support property to DistMixin to indicate the support of the probability distribution
• Add _check_support internal method that checks if a given value lies within the support of the distribution, and if not, shows a warning
• Set _type attribute for ContinuousDistMixin and DiscreteDistMixin to "continuous" and "discrete", respectively
• Set the support of all the distribution classes in wnb.dist module in their _support_ attribute

v0.1.7

What's Changed

Changelog v0.1.7:

• Add a parent mixin DistMixin for continuous and discrete distribution mixin
• Restructure ContinuousDistMixin and DiscreteDistMixin by inherting from DistMixin
• Add get_params() method to distribution mixins to return the distribution parameters
• Add a common __repr__ method to distribution mixins