boladjivinny / olpy Goto Github PK

While using the partial_fit method in the CW, the _update method raises an error while computing the covariance matrix.

Steps to reproduce

• Load a dataset
• Initializes a CW model
• Do a partial_fit using one data point

Traceback (most recent call last):
File "script.py", line 88, in
model = model.partial_fit(x, y, classes=[1, -1])
File "/home/boladji/project/pyenv/lib/python3.8/site-packages/olpy_boladjivinny-1.0.0-py3.8.egg/olpy/classifiers/__base.py", line 124, in partial_fit
self._update(x, y)
File "/home/boladji/project/pyenv/lib/python3.8/site-packages/olpy_boladjivinny-1.0.0-py3.8.egg/olpy/classifiers/cw.py", line 79, in _update
v_t = x @ np.diag(np.diag(self._sigma)) @ x.T
ValueError: matmul: Input operand 1 has a mismatch in its core dimension 0, with gufunc signature (n?,k),(k,m?)->(n?,m?) (size 1 is different from 123)

Vinny, this is really really awesome! I love your compliance with PEP-8 standards, use of poetry, code elegance, and neat documentation. I'll make some time this week to try it out to train some models for online learning!

Looking through the code, I realized that this package could take advantage of type hints just to make readability, understanding of the expected outputs, code linting and IDE integration even more easier.

As an example, the class olpy.preprocessing.labels.LabelEncoder has a fit(self, y) method. Adding type hints will yield:

import numpy as np

class LabelEncoder 
    
    def fit(self, y:np.ndarray) -> LabelEncoder:
        # method code goes here
        return self

Hope this helps! Would probably submit a few PRs to help with this after trying out the package. Cheers!